DataStax Enterprise 4.0

DataStax Enterprise 4.0
Documentation
July 1, 2015
©
2015 DataStax. All rights reserved.
Contents
Contents
About DataStax Enterprise............................................................................................. 7
Using the in-memory option...........................................................................................8
Upgrading....................................................................................................................... 11
Compression.................................................................................................................. 12
Installing..........................................................................................................................13
Installing on RHEL-based systems................................................................................................... 13
Installing on Debian-based systems................................................................................................. 14
Installing the binary tarball................................................................................................................ 16
Installing on SUSE............................................................................................................................ 17
Installing on cloud providers............................................................................................................. 18
Installing a DSE cluster on EC2............................................................................................ 18
Installing prior releases..................................................................................................................... 18
Security........................................................................................................................... 20
Security management........................................................................................................................20
Authenticating with Kerberos.............................................................................................................22
Creating Kerberos users.........................................................................................................24
Enabling and disabling Kerberos........................................................................................... 25
Using cqlsh with Kerberos......................................................................................................25
Client-to-node encryption...................................................................................................................25
Node-to-node encryption................................................................................................................... 26
Preparing server certificates..............................................................................................................26
Installing the cqlsh security packages.............................................................................................. 27
Running cqlsh......................................................................................................................... 28
Transparent data encryption............................................................................................................. 30
Encrypting data.......................................................................................................................31
Configuring encryption options............................................................................................... 32
Migrating encrypted tables..................................................................................................... 34
Configuring and using data auditing................................................................................................. 35
Formats of logs.......................................................................................................................37
Configuring auditing for a DSE Search/Solr cluster............................................................... 39
Configuring and using internal authentication...................................................................................40
Configuring internal authentication and authorization............................................................ 41
Changing the default superuser............................................................................................. 41
Enable internal security without downtime............................................................................. 42
Logging in with cqlsh..............................................................................................................42
Managing object permissions using internal authorization............................................................... 43
Configuring system_auth keyspace replication................................................................................. 44
Configuring firewall port access........................................................................................................ 44
2
Contents
DSE Analytics with Hadoop......................................................................................... 47
Getting started with Analytics and Hadoop in DataStax Enterprise..................................................47
Hadoop getting started tutorial............................................................................................... 48
Analytics node configuration...................................................................................................50
Changing the Hadoop log directory....................................................................................... 51
Portfolio Manager demo......................................................................................................... 52
Using the job tracker node................................................................................................................54
Setting the job tracker node...................................................................................................54
Using common hadoop commands........................................................................................55
Managing the job tracker using dsetool commands...............................................................55
Changing the job tracker client port....................................................................................... 56
About the Cassandra File System.................................................................................................... 57
Using the cfs-archive to store huge files.......................................................................................... 58
Using Hive......................................................................................................................................... 59
Running Hive.......................................................................................................................... 61
Browsing through Cassandra tables in Hive.......................................................................... 62
Creating or altering CQL data from Hive............................................................................... 62
Using a managed table to load local data............................................................................. 66
Using an external file system................................................................................................. 67
Example: Work with an unsupported data type..................................................................... 68
INSERT INTO SELECT statement.........................................................................................71
Example: Use a CQL composite partition key....................................................................... 72
Using CQL collections............................................................................................................ 73
Creating a Hive CQL output query........................................................................................ 75
Using a custom UDF.............................................................................................................. 76
Using pushdown predicates................................................................................................... 77
Using the Hive count function................................................................................................ 77
Handling schema changes..................................................................................................... 78
MapReduce performance tuning............................................................................................ 78
Starting the Hive server..........................................................................................................79
Using Beeline..........................................................................................................................80
Setting the Job Tracker node for Hive................................................................................... 80
Recreating Hive metadata after decommissioning a node.....................................................80
ODBC driver for Hive........................................................................................................................ 81
Configuring the driver............................................................................................................. 81
Using the driver...................................................................................................................... 82
Using Mahout.................................................................................................................................... 83
Using Mahout commands in DataStax Enterprise................................................................. 84
Using Pig........................................................................................................................................... 84
Running the Pig demo............................................................................................................85
Example: Save Pig relations from/to Cassandra....................................................................85
Example: Handle a compound primary key........................................................................... 87
Example: Explore library data................................................................................................ 90
Data access using storage handlers......................................................................................92
CQL data access....................................................................................................................94
CQL pushdown filter............................................................................................................... 96
Saving a Pig relation to Cassandra....................................................................................... 96
Creating a URL-encoded prepared statement....................................................................... 97
Sqoop.................................................................................................................................................98
About Sqoop........................................................................................................................... 98
Running the Sqoop demo...................................................................................................... 98
Validating import results in a cluster.................................................................................... 100
Migrating data to a Cassandra table....................................................................................100
About the generated Sqoop JAR file................................................................................... 100
3
Contents
Getting information about the sqoop command................................................................... 100
DSE Search with Solr..................................................................................................101
Getting started with Solr in DataStax Enterprise............................................................................ 101
Supported and unsupported features..............................................................................................102
Defining key Solr terms...................................................................................................................103
Installing Solr nodes........................................................................................................................ 104
Solr getting started tutorial.............................................................................................................. 104
Create a Cassandra table.................................................................................................... 105
Import data............................................................................................................................105
Create a search index.......................................................................................................... 105
Exploring the Solr Admin......................................................................................................106
Running a simple search......................................................................................................107
Running a faceted search.................................................................................................... 109
Solr HTTP API tutorial..........................................................................................................111
Configuring Solr...............................................................................................................................113
Mapping of Solr types.......................................................................................................... 113
Legacy mapping of Solr types..............................................................................................115
Configuring the Solr type mapping version.......................................................................... 116
Changing Solr Types............................................................................................................ 116
Configuring search components...........................................................................................116
Configuring multithreaded DocValuesFacets....................................................................... 117
Configuring the schema........................................................................................................117
Configuring the Solr library path.......................................................................................... 120
Configuring the Data Import Handler................................................................................... 120
Creating a Solr index...................................................................................................................... 122
Uploading the schema and configuration.............................................................................123
Creating a Solr core............................................................................................................. 123
Reloading a Solr core...........................................................................................................123
Rebuilding an index using the UI......................................................................................... 124
Checking indexing status......................................................................................................124
Adding and viewing index resources................................................................................... 125
Using DSE Search/Solr................................................................................................................... 125
Inserting, indexing, and searching data............................................................................... 127
Example: Using a CQL collection set.................................................................................. 127
Inserting/updating data using the Solr HTTP API................................................................ 128
Using dynamic fields.............................................................................................................129
Deleting Solr data................................................................................................................. 131
Using copy fields.................................................................................................................. 131
Viewing the Solr core status................................................................................................ 135
Querying Solr data.......................................................................................................................... 140
Using SolrJ and other Solr clients........................................................................................140
Shard selection..................................................................................................................... 140
Using the ShardRouter Mbean.............................................................................................140
Using the Solr HTTP API..................................................................................................... 141
Delete by id.......................................................................................................................... 141
Joining cores.........................................................................................................................142
Tracing Solr HTTP requests.................................................................................................145
Limiting columns indexed and returned by a query............................................................. 146
Querying multiple tables....................................................................................................... 146
Querying using autocomplete/spellcheck............................................................................. 146
Using CQL............................................................................................................................ 147
Using the ExtendedDisMax query parser.............................................................................148
Capacity planning............................................................................................................................ 149
Mixing workloads in a cluster..........................................................................................................150
4
Contents
Common operations........................................................................................................................ 152
Handling inconsistencies in query results............................................................................ 152
Adding, decommissioning, repairing a node........................................................................ 152
Shuffling shards to balance the load....................................................................................153
Managing the location of Solr data...................................................................................... 154
Solr log messages................................................................................................................ 154
Changing the Solr connector port........................................................................................ 154
Securing a Solr cluster......................................................................................................... 155
Fast repair.............................................................................................................................155
Excluding hosts from Solr-distributed queries...................................................................... 156
Expiring a DSE Search column............................................................................................156
Changing the HTTP interface to Apache JServe Protocol................................................... 158
Shard transport options for DSE Search/Solr communications............................................158
Tuning DSE Search performance................................................................................................... 159
Commit and query latency MBeans..................................................................................... 159
Using table compression...................................................................................................... 164
Configuring the update handler and autoSoftCommit.......................................................... 164
Changing the stack size and memtable space.................................................................... 165
Managing the consistency level........................................................................................... 165
Configuring the available indexing threads.......................................................................... 165
Managing caching.................................................................................................................166
Tuning index size and range query speed...........................................................................166
Increasing read performance by adding replicas................................................................. 167
Changing the replication factor for a Solr keyspace............................................................ 168
Configuring re-indexing.........................................................................................................168
DSE Search/Solr versus Open Source Solr................................................................................... 169
Request processing and data transformation................................................................................. 170
API for transforming Cassandra/Solr data........................................................................... 171
FIT reference implementation...............................................................................................171
Interface for custom field types............................................................................................ 173
Deploying...................................................................................................................... 175
Production deployment planning..................................................................................................... 175
Configuring replication..................................................................................................................... 175
Single data center deployment....................................................................................................... 179
Multiple data center deployment..................................................................................................... 181
Single-token architecture deployment............................................................................................. 184
Calculating tokens........................................................................................................................... 186
Expanding a DataStax AMI cluster................................................................................................. 188
DataStax Enterprise tools........................................................................................... 189
The dse commands.........................................................................................................................189
The dsetool......................................................................................................................................190
Configuring the disk health checker................................................................................................192
Pre-flight check and yaml_diff tools................................................................................................ 192
Using the Cassandra bulk loader in a secure environment............................................................ 193
Moving data to/from other databases....................................................................... 195
Reference...................................................................................................................... 196
Installing glibc on Oracle Linux....................................................................................................... 196
Tarball file locations........................................................................................................................ 196
Package file locations..................................................................................................................... 197
5
Contents
Configuration (dse.yaml)..................................................................................................................199
Starting and stopping DataStax Enterprise.....................................................................................200
Starting as a service.............................................................................................................201
Starting as a stand-alone process....................................................................................... 201
Stopping a node................................................................................................................... 202
Verify DataStax Enterprise is running..............................................................................202
Troubleshooting............................................................................................................................... 202
Cassandra Log4j appender............................................................................................................. 203
Log4j search demo............................................................................................................... 206
Release notes...............................................................................................................208
Using the docs.............................................................................................................219
6
About DataStax Enterprise
About DataStax Enterprise
DataStax Enterprise is a big data platform built on Apache Cassandra that manages real-time, analytics,
and enterprise search data. DataStax Enterprise leverages Cassandra, Apache Hadoop, and Apache Solr
to shift your focus from the data infrastructure to using your data strategically.
DataStax Enterprise is a big data platform built on Apache Cassandra that manages real-time, analytics,
and enterprise search data. DataStax Enterprise leverages Cassandra, Apache Hadoop, and Apache Solr
to shift your focus from the data infrastructure to using your data strategically, as described in the DataStax
Enterprise overview.
New features
•
In-memory option
•
In certain applications, such as ad bidding, response time is critical. A few milliseconds advantage
over another application in responding to a bid request can determine which advertiser gets the
placement. DataStax Enterprise 4.0 provides the in-memory option for Cassandra to accommodate
these applications.
Cassandra 2.0.1-2.0.3 enhancements, including:
•
•
•
•
•
•
•
•
•
• Lightweight transactions, performance enhancements
• Column aliases
Upgraded Hive, Pig, Solr, Sqoop, and Mahout components
Optional TCP-based Solr communications that provide lower latency, improved throughput, and
reduced resource consumption
Support for Solr custom field types
New mbeans for obtaining Solr commit and query latency information
Support for new Hive 0.12 data types date, varchar, and decimal
Support for the Hive INSERT INTO SELECT statement to copy data from one table and inserting it into
another
Improved performance of Cassandra File System (CassandraFS) compaction on analytics/Hadoop
nodes
Capabilities for configuring the Disk Health Checker
Performance enhancements reading MapReduce files
7
Using the in-memory option
Using the in-memory option
DataStax Enterprise includes the in-memory option for storing data to and accessing data from memory
exclusively.
DataStax Enterprise includes the in-memory option for storing data to and accessing data from memory
exclusively. No disk I/O occurs. Consider using the in-memory option for storing a modest amount of data,
mostly composed of overwrites, such as an application for mirroring stock exchange data. Only the prices
fluctuate greatly while the keys for the data remain relatively constant. Generally, the table you design for
use in-memory should have the following characteristics:
•
•
•
Store a small amount of data
Experience a workload that is mostly overwrites
Be heavily trafficked
Check performance metrics using OpsCenter, for example, before and after using the in-memory option.
Limitation
Currently, the in-memory option uses memory in the Java heap. Manage available memory carefully. Use
the dsetool inmemorystatus command to get the size, capacity, and percentage of memory in MB
used by a table. Bytes are truncated. For example:
$ bin/dsetool inmemorystatus ks1 users
Keyspace
ColumnFamily
ks1
users
Size
0MB
Capacity
1MB
Usage
52%
Creating a table using the in-memory option
In CQL, to create a table that uses the in-memory option, add a CQL directive to the CREATE TABLE
statement. Use the compaction directive in the statement to specify the MemoryOnlyStrategy class and
size_limit_in_mb property, which limits the amount of data that the table can accommodate.
CREATE TABLE users (
uid text,
fname text,
lname text,
PRIMARY KEY (uid)
) WITH compaction= { 'class': 'MemoryOnlyStrategy', 'size_limit_in_mb': 1 }
AND caching = 'NONE';
To enable metered flushing, configure the memtable_flush_period_in_ms using the CREATE TABLE or
ALTER TABLE statement.
Altering an on-disk table
Use the ALTER TABLE statement to change a traditional table to one that uses the in-memory option,
or vice versa. For example, suppose you have a traditional table named emp. Using the DESCRIBE
command, you can see that the table is a traditional table by the absence of a line in the output that looks
something like this:
compaction={'size_limit_in_mb': '1', 'class': 'MemoryOnlyStrategy'} >
Alter the emp table to use the in-memory option and, as a best practice, disable caching:
ALTER TABLE emp WITH compaction =
{ 'class': 'MemoryOnlyStrategy', 'size_limit_in_mb': 1 }
AND caching = 'NONE';
8
Using the in-memory option
Limiting the size of tables
The size_limit_in_mb property is a required property of the in-memory option schema that you configure
using CREATE TABLE or ALTER TABLE. Valid values are 1 - 1024, which limits tables in memory to 1GB
(1024MB) per node. It is possible, but not recommended, to create multiple 1GB tables, but no single table
can exceed 1GB per node. For example, the total space you can allocate to a table in memory is 1GB *
Nodes / replication factor; therefore, this configuration in a 10 node cluster can accommodate 5GB of data
distributed over the cluster:
•
•
size_limit_in_mb=1024
replication factor = 2
Disabling key caching
DataStax recommends disabling caching on tables configured to use the in-memory option. An error is
logged if key caching is not disabled. Enabling row caching, on the other hand, causes an error condition.
To disable both types of caching, set the table caching property to NONE.
ALTER TABLE users WITH caching = 'NONE';
Managing available memory
Running in a distributed environment, DataStax Enterprise cannot prevent you from adding excessive data
that exceeds the available memory. Differences in the data size from node to node that might exist make
such prevention impossible. It is the Cassandra administrator's responsibility to manage available memory
carefully.
Failure to manage available memory when using the in-memory option results in an error message that
looks something like this when capacity is exceeded:
SEVERE: java.util.concurrent.ExecutionException:
com.datastax.driver.core.exceptions.UnavailableException: Not enough replica
available for query at consistency ONE (1 required but only 0 alive)
SEVERE: java.util.concurrent.ExecutionException:
com.datastax.driver.core.exceptions.NoHostAvailableException: All host(s)
tried for query failed (tried: /10.53.120.13 (null), abc.com/10.53.122.15
(Timeout during read), abc.com/10.53.120.18 (null))
.
.
.
Checking available memory
Cassandra does not hold any locks on data while running requests, so concurrent write requests might
exceed the size_limit_in_mb a bit. Cassandra provides the AllMemtablesDataSize metric to check
available memory, so you can ensure that you have more available memory for a table than the size limit
allows. Use OpsCenter or JMX to check the AllMemtablesDataSize metric to determine available memory.
As mentioned previously, memtables flushes do not reduce the size of in-memory data.
Checking table properties
In cqlsh, use the DESCRIBE command to view table properties.
cqlsh> DESCRIBE TABLE users;
The output includes the size limit of the table data, size_limit_in_mb and whether or not the table uses the
in-memory option:
CREATE TABLE users (
uid text PRIMARY KEY,
fname text,
lname text
) WITH
bloom_filter_fp_chance=0.010000 AND
caching='KEYS_ONLY' AND
9
Using the in-memory option
comment='' AND
dclocal_read_repair_chance=0.000000 AND
gc_grace_seconds=432000 AND
read_repair_chance=0.100000 AND
replicate_on_write='true' AND
populate_io_cache_on_flush='false' AND
compaction={'size_limit_in_mb': '1', 'class': 'MemoryOnlyStrategy'} AND
compression={'sstable_compression': 'LZ4Compressor'} AND
caching = 'NONE';
Overwriting data best practice
Overwrite data in memory using CQL insert or update operations. Overwriting in-memory data takes
advantage of the memory capacity you have.
Backing up and restoring data
The procedure for backing up and restoring data is the same for in-memory and on-disk data. During the
snapshot process, Cassandra flushes data to disk, and then creates hard links to the backup-up SSTable
files for each keyspace in another named directory.
Flushing data to disk
To enable flushing to disk of the memtable data, change the default setting of the
memtable_flush_period_in_ms table property from 0 (disable) to a higher number, such as every hour
(3600 seconds). When the memtable flush period expires, Cassandra writes the contents of the memtable
to disk, purges the data in the commit log. The size of in-memory data is not affected by flushing. When
Cassandra flushes data in tables using the in-memory option to disk, new SSTables replace the old
ones. When Cassandra flushes data to disk in tables that are not in-memory tables, old SSTables are not
replaced.
Flushing data to disk does not remove in-memory data from the heap, as previously mentioned.
To automatically flush data to disk, configure the memtable_flush_period_in_ms using the CREATE
TABLE or ALTER TABLE command. For example, configure the users_flushed table to flush the memtable
every 3600ms.
CREATE TABLE users_flushed (
uid text,
fname text,
lname text,
PRIMARY KEY (uid)
) WITH compaction={'class': 'MemoryOnlyStrategy', 'size_limit_in_mb': 1}
AND memtable_flush_period_in_ms = 3600 AND caching = 'NONE';
Alternatively, you can flush data to disk manually. To manually flush data to disk, use the nodetool flush
command. For example, in the bin directory, flush the data from mykeyspace and mytable:
nodetool flush mykeyspace mytable
The nodetool flush command performs the operation on the current node and results in the following
background operations:
•
•
Creates a new SSTable
Deletes the commit logs that refer to data in the flushed memtables
To save time, flushing data to disk is recommended before backing up in-memory data.
10
Upgrading DataStax Enterprise
Upgrading DataStax Enterprise
See the DataStax Upgrade Guide.
See Upgrading Datastax Enterprise in the DataStax Upgrade Guide.
11
Compression
Compression
Configure data compression on a per-table basis to optimize performance of read-dominated tasks.
The default compression for all data in DataStax Enterprise 4.0.4 is LZ4.
The default compression in Cassandra 2.0.5 changed from Snappy to LZ4. DataStax Enterprise 4.0 4.0.3 uses LZ4 to compress real-time Cassandra and Solr nodes. The Snappy compressor remains the
default for data stored in CassandraFS. For example, if you put a text file on the CassandraFS using the
Hadoop shell command, compression is Snappy. If you create a CQL table on an analytics/Hadoop node,
the compression is LZ4.
LZ4 is fastest type of compression available in Cassandra, followed by Snappy, then by Deflate. Search
nodes typically engage in read-dominated tasks, so maximizing storage capacity of nodes, reducing
the volume of data on disk, and limiting disk I/O can improve performance. You can configure data
compression on a per-table basis to optimize performance of read-dominated tasks.
You can change compression options using ALTER TABLE. You can configure the compression chunk
size for read/write access patterns and the average size of rows in the table.
12
Installing
Installing
DataStax Enterprise installation methods include GUI or text mode, unattended command line or properties
file, YUM and APT repository, and binary tarball.
Installing DataStax Enterprise using Yum repositories
Install DataStax Enterprise and OpsCenter using Yum repositories on RHEL-based systems.
Note: To install on SUSE, use the binary tarball installation.
For a complete list of supported platforms, see DataStax Enterprise Supported Platforms.
Before you begin
•
•
•
•
•
•
•
•
Yum Package Management application installed.
Root or sudo access to the install machine.
Latest version of Oracle Java SE Runtime Environment 7. See Installing the Oracle JRE.
Python 2.6+ (needed if installing OpsCenter).
Java Native Access (JNA) is required for production installations. See Installing the JNA.
If installing on a 64-bit Oracle Linux distribution, first install the 32-bit versions of glibc libraries.
If installing OpsCenter on a RHEL 5.x/CentOS 5.x machine, make sure that the EPEL (Extra Packages
for Enterprise Linux) are installed. See Installing EPEL on CentOS 5.x or RHEL 5.x.
Some RedHat-compatible distributions do not contain the Linux Standard Base Core module (redhatlsb-core) by default. If your distribution does not have this package, you must install it.
Also see Recommended production settings and the DataStax Enterprise Reference Architecture white
paper.
About this task
The packaged releases create a cassandra user. When starting DataStax Enterprise as a service, the
Cassandra and Hadoop tracker services run as this user. The service initialization script is located in /
etc/init.d/dse. Run levels are not set by the package.
Procedure
These steps install DataStax Enterprise. After installing, you must configure and start DataStax Enterprise.
In a terminal window:
1. Check which version of Java is installed:
$ java -version
Use the latest version of Oracle Java 7 on all nodes.
2. Add the DataStax Yum repository to a file called /etc/yum.repos.d/datastax.repo
[datastax]
name = DataStax Repo for DataStax Enterprise
baseurl=http://username:[email protected]/enterprise
enabled=1
gpgcheck=0
where username and password are the DataStax account credentials from your registration
confirmation email.
13
Installing
3. Install DataStax Enterprise:
$ sudo yum -y install dse-full-version-1
Note:
dse-full installs DataStax Enterprise and the DataStax Agent.
dse-full opscenter installs DataStax Enterprise, DataStax Agent, and OpsCenter
(Optional)
For example:
$ sudo yum -y install dse-full-4.0.3-1
For production installations, DataStax recommends installing the OpsCenter separate from the cluster.
See the OpsCenter documentation.
Removing the datastax-agent package also removes the DataStax Enterprise package.
Results
DataStax Enterprise is ready for configuration.
What to do next
•
•
•
•
•
Set the configuration properties on each node in the cluster for single or multiple data center
deployment.
Configure the heap dump directory to avoid server crashes.
Start DataStax Enterprise.
Configuration file locations.
During normal use, Yum creates a cache of metadata and packages. To clean all cached files from any
enabled repository run:
$ yum clean all
Installing DataStax Enterprise on Debian-based systems
Install DataStax Enterprise and OpsCenter using APT repositories on Debian-based systems.
For a complete list of supported platforms, see DataStax Enterprise Supported Platforms.
Before you begin
•
•
•
•
•
•
Aptitude Package Management (APT) application installed.
Root or sudo access to the install machine.
Latest version of Oracle Java SE Runtime Environment 7. See Installing the Oracle JRE.
Python 2.6+ (needed if installing OpsCenter).
Java Native Access (JNA) is required for production installations. See Installing the JNA.
If you are using Ubuntu 10.04 LTS, you must update to JNA 3.4, as described in Installing the JNA on
Debian or Ubuntu systems.
Also see Recommended production settings and the DataStax Enterprise Reference Architecture white
paper.
About this task
The packaged releases create a cassandra user. When starting DataStax Enterprise as a service, the
Cassandra and Hadoop tracker services run as this user. The service initialization script is located in /
etc/init.d/dse. Run levels are not set by the package.
14
Installing
Procedure
These steps install DataStax Enterprise. After installing, you must configure and start DataStax Enterprise.
In a terminal window:
1. Check which version of Java is installed:
$ java -version
Use the latest version of Oracle Java 7 on all nodes.
2. Add a DataStax repository file called /etc/apt/sources.list.d/datastax.sources.list:
$ echo "deb http://username:[email protected]/enterprise stable
main" | sudo tee -a /etc/apt/sources.list.d/datastax.sources.list
where username and password are the DataStax account credentials from your registration
confirmation email.
3. Add the DataStax repository key:
$ curl -L https://debian.datastax.com/debian/repo_key | sudo apt-key add Note: If you have trouble adding the key, use http instead of https.
4. Install DataStax Enterprise:
$ sudo apt-get update
$ sudo apt-get install dse-full=version-1 dse=version-1 dse-hive=version-1
dse-pig=version-1 dse-demos=version-1 dse-libsolr=version-1 dselibtomcat=version-1 dse-libsqoop=version-1 dse-liblog4j=version-1
dse-libmahout=version-1 dse-libhadoop-native=version-1 dselibcassandra=version-1 dse-libhive=version-1 dse-libpig=version-1 dselibhadoop=version-1
Note:
dse-full installs DataStax Enterprise and the DataStax Agent.
dse-full opscenter installs DataStax Enterprise, DataStax Agent, and OpsCenter
(Optional)
For example:
$ sudo apt-get update
$ sudo apt-get install dse-full=4.0.3-1 dse=4.0.3-1 dse-hive=4.0.3-1 dsepig=4.0.3-1 dse-demos=4.0.3-1 dse-libsolr=4.0.3-1 dse-libtomcat=4.0.3-1
dse-libsqoop=4.0.3-1 dse-liblog4j=4.0.3-1 dse-libmahout=4.0.3-1 dselibhadoop-native=4.0.3-1 dse-libcassandra=4.0.3-1 dse-libhive=4.0.3-1 dselibpig=4.0.3-1 dse-libhadoop=4.0.3-1
For production installations, DataStax recommends installing the OpsCenter separate from the cluster.
See the OpsCenter documentation.
Removing the datastax-agent package also removes the DataStax Enterprise package.
Results
DataStax Enterprise is ready for configuration.
What to do next
•
Set the configuration properties on each node in the cluster for single or multiple data center
deployment.
15
Installing
•
•
•
Configure the heap dump directory to avoid server crashes.
Start DataStax Enterprise.
Configuration file locations.
Installing DataStax Enterprise on any Linux-based platform or Mac OS X
Install DataStax Enterprise on any Linux-based platform, including 32-bit platforms.
About this task
Use this install method for Mac OS X and platforms without package support, or if you do not have or want
a root installation.
Before you begin
•
•
•
•
Latest version of Oracle Java SE Runtime Environment 7. See Installing the Oracle JRE.
Java Native Access (JNA) is required for production installations. See Installing the JNA.
If you are using Ubuntu 10.04 LTS, you must update to JNA 3.4, as described in Installing the JNA on
Debian or Ubuntu systems.
If you are using an older RHEL-based Linux distribution, such as CentOS-5, you may need to replace
theSnappy compression/decompression library; see the Release Notes.
Also see Recommended production settings and the DataStax Enterprise Reference Architecture white
paper.
About this task
The binary tarball runs as a stand-alone process.
Procedure
These steps install DataStax Enterprise. After installing, you must configure and start DataStax Enterprise.
In a terminal window:
1. Check which version of Java is installed:
$ java -version
Use the latest version of Oracle Java 7 on all nodes.
2. Download the tarball from the Download DataStax Enterprise page.
You will need your DataStax account credentials from your registration confirmation email.
3. Unpack the distribution:
$ tar -xzvf dse-4.0.x.tar.gz
Note:
For production installations, DataStax recommends installing the OpsCenter separate from the
cluster. See the OpsCenter documentation.
Removing the datastax-agent package also removes the DataStax Enterprise package.
4. When DataStax Enterprise is started, it installs files into the /var/lib/cassandra and /var/log/
cassandra directories. If you do not have root access to the default directories, ensure that you have
write access:
$ sudo mkdir -p /var/lib/cassandra; sudo chown -R
cassandra
16
$USER: $GROUP /var/lib/
Installing
$ sudo mkdir -p /var/log/cassandra; sudo chown -R $USER: $GROUP /var/log/
cassandra
5. (Optional) If you do not want to use the default data and logging directories, you can define your own
directory locations:
a) Make the directories for data and logging directories:
$ mkdir install_location/dse-data
$ cd dse-data
$ mkdir commitlog
$ mkdir saved_caches
b) Go the directory containing the cassandra.yaml file:
$ cd install_location/resources/cassandra/conf
c) Edit the following lines in the cassandra.yaml file:
data_file_directories: install_location/dse-data
commitlog_directory: install_location/dse-data/commitlog
saved_caches_directory: install_location/dse-data/saved_caches
Results
DataStax Enterprise is ready for configuration.
What to do next
•
•
•
•
Set the configuration properties on each node in the cluster for single or multiple data center
deployment.
Configure the heap dump directory to avoid server crashes.
Start DataStax Enterprise.
Configuration file locations.
Installing the DataStax Enterprise tarball on SUSE Enterprise
DataStax provides a binary tarball distribution for installing DataStax Enterprise on SUSE Linux.
About this task
DataStax provides a binary tarball distribution for installing on SUSE Linux. For a complete list of supported
platforms, see DataStax Enterprise Supported Platforms.
Before you begin
DataStax account credentials from your registration confirmation email.
Also see Recommended production settings and the DataStax Enterprise Reference Architecture white
paper.
Also see Recommended production settings.
Procedure
1. Install DataStax Enterprise using the Installing the binary tarball instructions.
2. Set up DataStax Enterprise as described in the Deploying section.
3. Start DataStax Enterprise and the DataStax Agent on each node as described in Starting and stopping
DataStax Enterprise.
4. To install OpsCenter, use the OpsCenter Tarball Distribution instructions.
17
Installing
Installing on cloud providers
Install on Amazon EC2 or HP cloud.
Installing a DataStax Enterprise cluster on Amazon EC2
Installing a DataStax Enterprise cluster on Amazon EC2.
About this task
For instructions on installing the DataStax AMI (Amazon Machine Image), see the latest AMI
documentation.
Installing prior releases of DataStax Enterprise
Steps for installing the same version as other nodes in your cluster.
About this task
DataStax provides binary tarball and packaged releases for installing earlier releases (2.2.x upwards ) of
DataStax Enterprise.
Note: You must use Oracle JRE 6, not 7 for releases before DataStax Enterprise 3.1. These earlier
releases do not support JRE 7.
Installing from the binary tarball
Download the tarball from the Download DataStax Enterprise page and follow the install instructions in the
relevant documentation:
•
•
DataStax Enterprise 2.2.x tarball install documentation
DataStax Enterprise 3.0.x tarball install documentation
Installing the packages on RHEL-based or Debian-based platforms
Follow the install instructions in the relevant documentation and specify the specific version in the install
command:
•
•
DataStax Enterprise 2.2.x install documentation
DataStax Enterprise 3.0.x install documentation
RHEL-based platforms
Format:
$ sudo yum install dse-full-version-1
Example:
$ sudo yum install dse-full-3.0.3-1
Debian-based platforms
Format:
$ sudo apt-get install dse-full=version-1 dse=version-1 dse-hive=version-1
dse-pig=version-1 dse-demos=version-1 dse-libsolr=version-1 dselibtomcat=version-1 dse-libsqoop=version-1 dse-liblog4j=version-1 dselibmahout=version-1 dse-libhadoop-native=version-1 dse-libcassandra=version-1
dse-libhive=version-1 dse-libpig=version-1 dse-libhadoop=version-1
Example:
18
Installing
$ sudo apt-get install dse-full=3.0.3-1 dse=3.0.3-1 dse-hive=3.0.3-1 dsepig=3.0.3-1 dse-demos=3.0.3-1 dse-libsolr=3.0.3-1 dse-libtomcat=3.0.3-1 dselibsqoop=3.0.3-1 dse-liblog4j=3.0.3-1 dse-libmahout=3.0.3-1 dse-libhadoopnative=3.0.3-1 dse-libcassandra=3.0.3-1 dse-libhive=3.0.3-1 dse-libpig=3.0.3-1
dse-libhadoop=3.0.3-1
19
Security
Security
Managing security in DataStax Enterprise including authentication, encryption, auditing, permissions, and
configuration.
Security management
DataStax Enterprise includes advanced data protection for enterprise-grade databases including internal
authentication, object permissions, encryption, Kerberos authentication, and data auditing.
About this task
DataStax Enterprise 4.0 includes advanced data protection for enterprise-grade databases:
•
•
•
•
•
•
Internal authentication using login accounts and passwords
Managing object permissions using internal authorization based on the GRANT/REVOKE paradigm
Client-to-node encryption using SSL for data going from the client to the Cassandra cluster
Node to node encryption using SSL for data between nodes
Kerberos authentication for nodes to communication over a non-secure network by proving their identity
to one another in a secure manner using tickets
Configuring and using data auditing for an administrator to use to create detailed audit trails of cluster
activity
Transparent data encryption that transparently encodes data flushed from the memtable in system
memory to the SSTables on disk (at rest data), making the at rest data unreadable by unauthorized
users
•
The TCP-communications layer for Solr supports client-to-node and node-to-node encryption using SSL,
but does not support Kerberos.
The DataStax Java Driver 2.0 and DataStax C# Driver, available on the DataStax web site, enables
Kerberos support and also SSL for client/server communication.
Limitations
Assuming you configure security features, this table describes exactly which data is secured (or not) based
on the workload type: real-time Cassandra (DSE/Cassandra), analytics (Hadoop), and DSE/Search (Solr).
Feature
DSE/Cassandra
Hadoop
Solr
Internal authentication
Yes
Yes
No
Object permission
management
Yes
Partial [1]
Partial [1]
Client to node encryption Yes [2]
Yes [3]
Yes [4]
Kerberos authentication
Yes [5]
Yes
Yes
Transparent data
encryption
Yes [6]
Yes
Partial [7]
Data auditing
Yes
Partial [8]
Full
[1] Permissions to access objects stored in Cassandra are checked. The Solr cache and indexes and the
Hadoop cache are not under control of Cassandra, and therefore are not checked. You can, however, set
up permission checks to occur on tables that store Hadoop or Solr data.
20
Security
[2] The inter-node gossip protocol is protected using SSL.
[3] The Thrift interface between Hadoop and the Cassandra File System (CFS) is SSL-protected. Intertracker communication is Kerberos authenticated, but not SSL secured. Hadoop access to Cassandra is
SSL- and Kerberos-protected.
[4] HTTP access to the DSE Search/Solr data is protected using SSL. Node-to-node encryption using SSL
protects internal Solr communication.
[5] The inter-node gossip protocol is not authenticated using Kerberos. Node-to-node encryption using SSL
can be used.
[6] Cassandra commit log data is not encrypted, only at rest data is encrypted.
[7] Data in DSE/Search Solr tables is encrypted by Cassandra. Encryption has a slight performance
impact, but ensures the encryption of original documents after Cassandra permanently stores the
documents on disk. However, Solr cache data and Solr index data (metadata) is not encrypted.
[8] Hadoop data auditing is done at the Cassandra access level, so requests to access Cassandra data is
audited. Full Solr auditing is done. Node-to-node encryption using SSL protects communication over internode gossip protocol.
Using Kerberos and SSL at the same time
Both the Kerberos and SSL libraries provide authentication, encryption, and integrity protection:
•
•
•
Kerberos - If you enable Kerberos authentication, integrity protection is also enabled. However, you can
enable integrity protection without encryption.
SSL - If you use SSL, authentication, integrity protection, and encryption are all enabled or disabled.
Kerberos and SSL - It is possible to enable both Kerberos authentication and SSL together. However,
this causes some overlap because authentication is performed twice by two different schemes:
Kerberos authentication and certificates through SSL. DataStax recommends choosing one and using it
for both encryption and authentication. These settings are described in the dse.yaml configuration file.
Securing DSE Search services
The security table summarizes the security features of DSE Search/Solr and other integrated components.
DSE Search data is completely or partially secured by using these DataStax Enterprise security features:
•
Managing object permissions using internal authorization
•
Access to Solr documents, excluding cached data, can be limited to users who have been granted
access permissions. Permission management also secures tables used to store Solr data.
Transparent data encryption
•
Data at rest in Cassandra tables, excluding cached and Solr-indexed data, can be encrypted.
Encryption occurs on the Cassandra side and impacts performance slightly.
Client-to-node encryption
•
You can encrypt HTTP access to Solr data and internal, node-to-node Solr communication using SSL.
Enable SSL node-to-node encryption on the Solr node by setting encryption options in the dse.yaml
file as described in Client-to-node encryption.
Kerberos authentication
You can authenticate DSE Search users through Kerberos authentication using Simple and Protected
GSSAPI Negotiation Mechanism (SPNEGO).
You can also use HTTP Basic Authentication, but this is not recommended.
sstableloader security options
In DataStax Enterprise 4.0, the procedure for securing sstableloader has changed slightly from previous
releases.
21
Security
Authenticating a cluster with Kerberos
DataStax Enterprise authentication with Kerberos protocol uses tickets to prove identity for nodes that
communicate over non-secure networks.
About this task
This section provides information about configuring security for a DataStax Enterprise (DSE) cluster using
Kerberos.
Kerberos is a computer network authentication protocol that allows nodes communicating over a nonsecure network to prove their identity to one another in a secure manner using tickets. This section does
not provide detailed information on installing and setting up Kerberos. For this information, see the MIT
Kerberos Consortium.
Note: When using Kerberos security, you need to be aware of the scope of Kerberos tickets. Using
the su or sudo command leaves any existing credentials behind and requires you to re-authenticate
as that new user. If you encounter authentication issues, please ensure that you have a proper
Kerberos ticket.
For information about using Kerberos with SSL, see Using Kerberos and SSL at the same time.
About this task
Kerberos Recommendations
The following are general guidelines for setting up Kerberos:
•
•
•
•
•
•
Before installing DSE, set up your Kerberos servers.
Set up several machines as authentication servers (Key Distribution Center [KDC]). One will be the
primary or administration KDC, the others will be secondary.
Do not install the KDC servers on DSE nodes.
Set up firewalls on each KDC server.
Physically protect the KDC machines.
Secure the keytab files owned by the user running DSE. The file should be readable and writeable only
by the owner without permissions for any other user (chmod 0600).
AES-256 support
Because JCE-based products are restricted for export to certain countries by the U.S. Export
Administration Regulations, DataStax Enterprise does not ship with the Java Cryptography Extension
(JCE) Unlimited Strength Jurisdiction Policy. DataStax recommends installing the JCE Unlimited Strength
Jurisdiction Policy Files:
1. Go to the Oracle Java SE download page.
•
For Java 6, click Previous Releases > Java Platform Technologies > Java Cryptography
Extension (JCE) Unlimited Strength Jurisdiction Policy Files 6.
• For Java 7, under Additional Resources, download the Java Cryptography Extension (JCE)
Unlimited Strength Jurisdiction Policy Files.
2. Unzip the downloaded file.
3. Copy local_policy.jar and US_export_policy.jar to the $JAVA_HOME/jre/lib/security
directory overwriting the existing JARS.
If you choose not to use AES-256, you must remove the AES-256 settings as an allowed cypher for each
principal and regenerate the keys for the krbtgt principal. Remove AES-256 settings in one of the following
ways:
22
Security
•
•
If you have not created the principals, use the -e flag to specify encryption:salt type pairs. For example:
-e "arcfour-hmac:normal des3-hmac-sha1:normal". This method requires Kerberos 5-1.2 on the KDC.
If you have already created the principals, modify the Kerberos principals using the -e flag as described
above and then recreate the keytab file. This method requires Kerberos 5-1.2 on the KDC.
Alternately, you can modify the /etc/krb5kdc/kdc.conf file by removing any entries containing
aes256 from the supported_enctypes variable for the realm in which the DSE nodes are members.
Then change the keys for the krbtgt principal.
Note: If the KDC is used by other applications, changing the krbtgt principal's keys invalidates any
existing tickets. To prevent this, use the -keepold option when executing the change_password
command. For example: 'cpw -randkey krbtgt/krbtgt/[email protected]'
Securing DataStax Enterprise nodes
Do not upgrade DataStax Enterprise and set up Kerberos at the same time; see Security
Recommendations.
Procedure
Perform the following on every node:
1. Install the Kerberos client software.
2. If you are not using the JCE Unlimited Strength Jurisdiction Policy, make sure that your ticket granting
principal does not use AES-256 as described above.
3. Use Kerberos to generate one keytab file for each node:
kadmin -p username/admin
addprinc -randkey dse/FQDN
addprinc -randkey HTTP/FQDN
ktadd -k dse.keytab dse/FQDN
ktadd -k dse.keytab HTTP/FQDN
quit
•
•
-randkey creates a random password.
ktadd -k creates a keytab for the dse and HTTP principals; -k specifies the keytab file name. In
this example, the keytab entry is added to the dse.keytab file in the current directory.
4. In the cassandra.yaml configuration file, set the authenticator:
authenticator: com.datastax.bdp.cassandra.auth.KerberosAuthenticator
5. Change the replication strategy and default replication factor for the system_auth keyspace. See
Configuring system_auth keyspace replication.
DataStax recommends configuring system_auth keyspaces for fault tolerance (in case of failure). In a
multi-node cluster, if the node storing the user data goes down, using the default replication factor of 1
for the system_auth keyspace precludes logging into any secured node.
6. Set the DSE service principals, keytab location, and qop (Quality of Protection) in the dse.yaml
configuration file:
kerberos_options:
keytab: path_to_keytab/dse.keytab
service_principal: dse_user/[email protected]
http_principal: HTTP/[email protected]
qop: auth
•
Set the service_principal that the Cassandra and Hadoop processes run under. It must use the form
dse_user/[email protected], where dse_user is cassandra in package installs (the name of the user
running the service) and the name of the UNIX user that starts the service in binary installs. It must
be consistent everywhere: in the dse.yaml, present in the keytab, and in the cqlshrc file (where it
is separated into the service/hostname).
23
Security
•
•
•
•
•
Set REALM to the name of your Kerberos realm. In the Kerberos principal, REALM must be all
uppercase.
Leave _HOST as is. DSE automatically substitutes the FQDN (Fully Qualified Domain Name) of the
host where it runs. There must be credentials for this principal in the keytab file and readable by the
user that Cassandra runs as, usually cassandra.
The http_principal is used by the application container, which is tomcat, and used to run Solr.
The web server uses GSS-API mechanism (SPNEGO) to negotiate the GSSAPI security mechanism
(Kerberos). To set up password authentication for a DSE Search/Solr node, see Securing a Solr
cluster.
The keytab file must contain the credentials for both of the fully resolved principal names, which
replace _HOST with the FQDN of the host in the service_principal and http_principal
settings. The UNIX user running DSE must also have read permissions on the keytab.
The qop is a comma delimited list of Quality of Protection values that clients and servers can use for
each connection. The client can have multiple QOP values, while the server can have only a single
QOP value. The available settings are:
•
•
•
auth - authentication only [default]
auth-int - authentication plus integrity protection for all transmitted data
auth-conf - authentication plus integrity protection and encryption of all transmitted data
For example, if the realm name is foo.com and keytab file is in the resources/dse/conf
directory:
kerberos_options:
keytab: resources/dse/conf/dse.keytab
service_principal: cassandra/[email protected]
http_principal: HTTP/[email protected]
qop: auth
Be sure that the realm name is uppercase.
Creating Kerberos users
You can use password authentication or the cassand[email protected] Kerberos principal to create Kerberos
users.
About this task
DataStax Enterprise automatically creates a cassandra superuser, which you can authenticate as and use
cqlsh to create other users. Two methods are available:
•
Use password authentication:
1. In the cassandra.yaml file, set the authenticator to
org.apache.cassandra.auth.PasswordAuthenticator.
2. Start cqlsh and login using the superuser name and password:
$ ./cqlsh -u cassandra -p cassandra
3. Create the other Kerberos users, such as [email protected] Be sure to create at least one with
superuser privileges.
4. Remove the cassandra user. See DROP USER. This step is optional but highly recommended.
5. Re-enable Kerberos authorization in the cassandra.yaml file:
•
authenticator: com.datastax.bdp.cassandra.auth.KerberosAuthenticator
Use the [email protected] Kerberos principal:
1. As shown in step 6 in Authenticating a DataStax Enterprise cluster with Kerberos, create a
[email protected] Kerberos principal and turn on Kerberos authorization.
2. Log in and create the other Kerberos users. Be sure to create at least one with superuser privileges.
3. Remove the cassandra user. See DROP USER. This step is optional but highly recommended.
24
Security
Enabling and disabling Kerberos security
Turn Kerberos authorization on and off by changing the authenticator in cassandra.yaml.
After setting up Kerberos as described above, you can turn it on and off by changing the authenticator in
the cassandra.yaml file:
•
•
On: com.datastax.bdp.cassandra.auth.KerberosAuthenticator
Off: any other authenticator
Using cqlsh with Kerberos security
Install required packages to use cqlsh with Kerberos.
About this task
Install required packages to use cqlsh with Kerberos.
Procedure
To use cqlsh with Kerberos:
1. Install the python-kerberos and python-pure-sasl packages.
See Installing the cqlsh security packages.
2. Create a cqlshrc file in your ~/.cassandra or client program ~/.cassandra directory.
Client-to-node encryption
Client-to-node encryption protects data in flight from client machines to a database cluster.
About this task
Client-to-node encryption protects data in flight from client machines to a database cluster. It establishes
a secure channel between the client and the coordinator node. Unlike Kerberos, SSL is fully distributed
and does not require setting up a shared authentication service. For information about generating SSL
certificates, see Preparing server certificates.
SSL settings for DataStax Enterprise client-to-node encryption
About this task
To enable client-to-node SSL, set the client encryption options. Where you set them depends on the
version.
Procedure
1. Set the client encryption options using one of the two following scenarios.
Configure the client_encryption_options only in the cassandra.yaml file. If necessary, remove them from
the dse.yaml.
2. On each node, under client_encryption_options:
•
•
•
Enable encryption.
Set the paths to your .keystore and .truststore files.
Provide the passwords used when generating the keystore and truststore.
client_encryption_options:
enabled: true
keystore: resources/dse/conf/.keystore
keystore_password: keystore password
25
Security
store_type: JKS
truststore: resources/dse/conf/.truststore
truststore_password: truststore password
protocol: ssl
cipher_suites: [TLS_RSA_WITH_AES_128_CBC_SHA,
TLS_RSA_WITH_AES_256_CBC_SHA]
For information about using Kerberos with SSL, see Using Kerberos and SSL at the same time.
Note: Initializing Solr to support SSL encryption
When you enable SSL, it automatically enables the authentication/authorization filters in Solr
web.xml and configures an SSL connector in Tomcat. This means that you don't have to change
your web.xml or server.xml.
Node-to-node encryption
Node-to-node encryption protects data that is transferred between nodes in a cluster using SSL (Secure
Sockets Layer).
About this task
Node-to-node encryption protects data transferred between nodes in a cluster using SSL (Secure Sockets
Layer). For information about generating SSL certificates, see Preparing server certificates.
SSL settings for node-to-node encryption
To enable node-to-node SSL, you must set the encryption options in the cassandra.yaml file.
On each node, under encryption_options:
•
•
•
•
Enable the internode_encryption options (described below).
Set the appropriate paths to your .keystore and .truststore files.
Provide the required passwords. The passwords must match the passwords used when generating the
keystore and truststore.
To enable peer certificate authentication, set require_client_auth to true.
The available inter-node options are:
•
•
•
•
all
none
dc - Cassandra encrypts the traffic between the data centers.
rack - Cassandra encrypts the traffic between the racks.
encryption_options:
internode_encryption: internode_option
keystore: resources/dse/conf/.keystore
keystore_password: keystore password
truststore: resources/dse/conf/.truststore
truststore_password: truststore password
require_client_auth: true or false
Preparing server certificates
Generate SSL certificates for client-to-node encryptions or node-to-node encryption.
26
Security
About this task
This topic provides information about generating SSL certificates for client-to-node encryption or node-tonode encryption. If you generate the certificates for one type of encryption, you do not need to generate
them again for the other: the same certificates are used for both.
All nodes must have all the relevant SSL certificates on all nodes. A keystore contains private keys. The
truststore contains SSL certificates for each node and doesn't require signing by a trusted and recognized
public certification authority.
Procedure
To prepare server certificates:
1. Generate the private and public key pair for the nodes of the cluster leaving the key password the same
as the keystore password:
keytool -genkey -alias dse_node0 -keyalg RSA -keystore .keystore
2. Repeat the previous step on each node using a different alias for each one.
3. Export the public part of the certificate to a separate file and copy these certificates to all other nodes.
keytool -export -alias dse -file dsenode0.cer -keystore .keystore
4. Add the certificate of each node to the truststore of each node, so nodes can verify the identity of other
nodes.
A prompt for setting a password for the newly created truststore appears.
keytool -import -v -trustcacerts -alias dse_node0 -file dse_node0.cer keystore .truststore
keytool -import -v -trustcacerts -alias dse_node1 -file dse_node1.cer keystore .truststore
. . .
keytool -import -v -trustcacerts -alias dse_nodeN -file dse_nodeN.cer keystore .truststore
5. Make sure .keystore is readable only by the DSE daemon and not by any user of the system.
Installing the cqlsh security packages
Install packages to use cqlsh with a Kerberized cluster.
About this task
To use cqlsh with a Kerberized cluster, you must install the PyKerberos and python-pure-sasl packages.
The PyKerberos package is a high-level wrapper for Kerberos (GSSAPI) operations. The python-pure-sasl
package is a pure Python client-side SASL (Simple Authentication and Security Layer) implementation.
RHEL-based installs
From root:
1. Make sure that the DataStax repository has been added. See Installing DataStax Enterprise on RHELbased systems.
2. Check which version of Python is installed:
python -V
3. Add the Python module:
• # yum install python26-pure-sasl ## for Python 2.6.x
• # yum install python27-pure-sasl ## for Python 2.7.x
4. Add the Kerberos module:
•
# yum install python26-kerberos ## for Python 2.6.x
27
Security
•
# yum install python27-kerberos ## for Python 2.7.x
Debian-based installs
1. Make sure that the DataStax repository has been added. See Installing DataStax Enterprise on Debianbased systems.
2. Add the modules:
$ sudo apt-get install python-pure-sasl
This command installs both the Python and Kerberos modules.
Tarball installs
Attention: DataStax recommends using APT or Yum because installing the dependencies can be
difficult, time consuming, and requires a high level of Linux expertise.
1. Ensure all dependencies are properly installed:
• Debian-based systems: $ apt-cache show python-kerberos
• RHEL-based systems: $ yum deplist python-kerberos
2. Look at the Depends field and update your system to meet any dependencies, such as gcc, g++,
python-dev, and libkrb5-dev.
3. Download the PyKerberos tarball:
$ curl -OL http://username:[email protected]/enterprise/
kerberos-1.1.2+DSE1.tar.gz
4. Extract the tarball:
$ tar -xzf kerberos-1.1.2+DSE1.tar.gz
5. From the directory where you untarred PyKerberos:
$ python setup.py build
6. From the install directory:
$ python setup.py install
7. Download the pure-sasl module tarball:
$ curl -OL http://pypi.python.org/packages/source/p/pure-sasl/puresasl-0.1.3.tar.gz
8. Extract the tarball:
$ tar -xzf pure-sasl-0.1.3.tar.gz
9. From the install directory:
$ sudo python setup.py install
Running cqlsh
Sample files for Kerberos, SSL, and Kerboros and SSL.
To run cqlsh, you need to create a cqlshrc file in your ~/.cassandra directory. You cannot use cqlsh
when client certificate authentication is enabled (require_client_auth=true). Sample files are available in the
following directories:
•
•
Packaged installs: /usr/share/doc/dse-libcassandra
Tarball installs: install_location/resources/cassandra/conf
Kerberos example
[connection]
hostname = 192.168.1.2
port = 9160
factory = cqlshlib.kerberos.kerberos_transport_factory
28
Security
[kerberos]
hostname = cassandra01.example.com
service = cassandra
principal = bill/[email protected] ;; Optional.
qops = auth-conf ;; Optional, see the following paragraph.
[kerberos-hostnames] ;; Optional section, overrides default hostname in
[kerberos] section.
192.168.1.3 = cassandra01.example.com
192.168.1.4 = cassandra02.example.com
If qops is not specified the default (auth) is used. On the client side, the qops option is a comma-delimited
list of the QOP values allowed by the client for the connection. The client (cqlsh) value list must contain at
least one of the QOP values specified on the server. To clarify, the client can have multiple QOP values,
while the server can have only a single QOP value (specified in the dse.yaml).
The Kerberos hostname and service are mandatory settings and must be provided either in the
configuration file or as environment variables. The environment variables (KRB_HOST, KRB_SERVICE,
and KRB_PRINCIPAL) override any options set in this file. For more information about these settings,
see Securing DataStax Enterprise nodes. The hostname and service must match the values set in the
dse.yaml.
SSL example
[connection]
hostname = 127.0.0.1
port = 9160
factory = cqlshlib.ssl.ssl_transport_factory
[ssl]
certfile = ~/keys/cassandra.cert
validate = true ;; Optional, true by default.
[certfiles] ;; Optional section, overrides the default certfile in the [ssl]
section.
192.168.1.3 = ~/keys/cassandra01.cert
192.168.1.4 = ~/keys/cassandra02.cert
Note: When generating the certificate, be sure to set the CN to the hostname of the node.
You must create a pem key which is used in the cqlshrc file.
keytool -importkeystore -srckeystore .keystore -destkeystore user.p12 deststoretype PKCS12
openssl pkcs12 -in user.p12 -out user.pem -nodes
The pem key is required because the host in the certificate is compared to the host of the machine that it
is connected to. The SSL certificate must be provided either in the configuration file or as an environment
variable. The environment variables (SSL_CERTFILE and SSL_VALIDATE) override any options set in this
file.
Kerberos and SSL
For information about using Kerberos with SSL, see Using Kerberos and SSL at the same time.
The settings for using both Kerberos and SSL are a combination of the Kerberos and SSL sections in the
above examples, except the factory setting:
factory = cqlshlib.kerberos_ssl.kerberos_ssl_transport_factory
The supported environmental variables are KRB_HOST, KRB_SERVICE, KRB_PRINCIPAL,
SSL_CERTFILE, and SSL_VALIDATE variables.
29
Security
Transparent data encryption
Transparent data encryption (TDE) protects at rest data. TDE requires a secure local file system to be
effective.
About this task
Transparent data encryption (TDE) protects at rest data. At rest data is data that has been flushed from the
memtable in system memory to the SSTables on disk.
As shown in the diagram, data stored in the commit log is not encrypted. If you need commit log
encryption, store the commit log on an OS-level encrypted file system using Vormetric, for example. Data
can be encrypted using different algorithms, or you can choose not to encrypt data at all. SSTable data
files are immutable (they are not written to again after they have been flushed to disk). SSTables are
encrypted only once when they are written to disk.
The CassandraFS (Cassandra file system) is accessed as part of the Hadoop File System (HDFS) using
the configured authentication. If you encrypt the CassandraFS keyspace's sblocks and inode tables, all
CassandraFS data gets encrypted.
Limitations and recommendations
Data is not directly protected by TDE when accessed using the following utilities.
Utility
Reason Utility Is Not Encrypted
json2sstable
Operates directly on the sstables.
nodetool
Uses only JMX, so data is not accessed.
sstable2json
Operates directly on the sstables.
sstablekeys
Operates directly on the sstables.
sstableloader
Operates directly on the sstables.
sstablescrub
Operates directly on the sstables.
Compression and encryption introduce performance overhead.
In DataStax Enterprise 4.0.5, if you use transparent data encryption, you need to run Opscenter agent
under the same account as the DataStax Enterprise user.
30
Security
Requirements
TDE requires a secure local file system to be effective. The encryption certificates are stored locally;
therefore, an invasion of the local file system invalidates encryption.
Options
To get the full capabilities of TDE, download and install the Java Cryptography Extension (JCE), unzip the
jar files and place them under $JAVA_HOME/jre/lib/security. JCE-based products are restricted for
export to certain countries by the U.S. Export Administration Regulations.
Encrypting data
Data encryption uses a system key in the dse_system.encrypted_keys table.
The procedure for encrypting data changed in DataStax Enterprise 4.0.4. First, you use a new
dsetool command to create a system key. The system key is used to encrypt all data in the
dse_system.encrypted_keys table. Next, copy the system key you created to all nodes in the cluster. The
individual encryption keys for each table are stored in the dse_system.encrypted_keys table, encrypted
using the system key. The location of the system key is configured in the dse.yaml. The default location
should be fine for a package install, but for a tarball installation you need to configure the path to a
directory that you have permission to access.
Encrypting data in DataStax Enterprise 4.0.4 and later
1. Back up SSTables.
2. On a packaged installation, accept the default system_key_directory /etc/dse/conf. Go to the next
step to set permissions on the directory.
On a tarball installation, optionally change the directory on each node in the cluster from /etc/dse/conf to
another directory, or skip this step, and adjust permissions as described in the next step.
•
•
•
Navigate to install-directory/resources/dse/conf.
Open the dse.yaml file for editing.
Change the path of the system_key_directory to the path of a directory that you have permission to
access.
3. Set permissions on the system_key_directory to give rights to change the keytab file only to the user/
group running DataStax Enterprise. If JNA is installed, and the permissions are too open, DataStax
Enterprise makes the change automatically.
4. Create a system key using the dsetool createsystemkey command. For example:
$ dsetool createsystemkey 'AES/ECB/PKCS5Padding' 128 system_key
5. Copy the key and paste it to the location specified by the system_key_directory on each node in the
cluster.
6. Perform a rolling restart on the cluster.
7. Check that the dse_system keyspace and encrypted_keys table now exist.
cqlsh:mykeyspace> DESCRIBE KEYSPACES;
system
dse_system
mykeyspace
system_traces
cqlsh:mykeyspace> USE dse_system;
cqlsh:dse_system> DESCRIBE TABLES dse_system;
encrypted_keys
8. Set encryption options as you create a table or alter an existing table.
Tables are encrypted when Cassandra stores the tables on disk as SSTables.
9. Rewrite all SSTables using the specified encryption by running nodetool upgradesstables --include-allsstables.
31
Security
Encrypting data in DataStax Enterprise 4.0 - 4.0.3
1. Back up SSTables.
2. Set permissions so that only the user/group running DataStax Enterprise can change the keytab file. If
JNA is installed, JNA takes care of setting these permissions.
3. Ensure that the user encrypting data has been granted ALTER permission on the table containing the
data to be encrypted.
4. Set encryption options as you create a table or alter an existing table.
5. Rewrite all SSTables using nodetool upgradesstables --include-all-sstables.
Configuring encryption options
Configure encryption on a per table basis. Each node generates a separate key used for only that node’s
SSTables.
You designate encryption on a per table basis. When using encryption, each node generates a separate
key used for only that node’s SSTables. To encrypt data, first log in as the default superuser. For example.
$ cqlsh -u cassandra -p cassandra
The ALTER TABLE syntax for setting encryption options is the same as the syntax for setting data
compression options.
For example, to set compression options in the users table:
ALTER TABLE users
WITH compression =
{ 'sstable_compression' : 'DeflateCompressor',
'chunk_length_kb' : 64 };
To set encryption options in the users table, for example:
ALTER TABLE users
WITH compression =
{ 'sstable_compression' : 'Encryptor',
'cipher_algorithm' : 'AES/ECB/PKCS5Padding',
'secret_key_strength' : 128,
'chunk_length_kb' : 1 };
Designating data for encryption using ALTER TABLE doesn't encrypt existing SSTables, just new
SSTables that are generated. When setting up data to be encrypted, but not compressed, set the
chunk_length_kb option to the lowest possible value, 1, as shown in the previous example. Setting this
option to 1 improves read performance by limiting the data that needs to be decrypted for each read
operation to 1 KB.
Setting encryption and compression together
Encryption and compression occur locally, which is more performant than trying to accomplish these tasks
on the Cassandra-side. Encryption can be set together with compression using a single statement. The
single CQL statement in is:
ALTER TABLE users
WITH compression =
{ 'sstable_compression' : 'EncryptingSnappyCompressor',
'cipher_algorithm' : 'AES/ECB/PKCS5Padding',
'secret_key_strength' : 128,
, 'chunk_length_kb' : 128 };
Encryption/compression options and sub-options
Using encryption, your application can read and write to SSTables that use different encryption algorithms
or no encryption at all. Using different encryption algorithms to encrypt SSTable data is similar to using
different compression algorithms to compress data. This section lists the options and sub-options.
32
Security
The high-level container option for encryption and/or compression used in the ALTER TABLE statement
are:
•
•
•
•
•
•
Encryptor
EncryptingDeflateCompressor
EncryptingSnappyCompressor
DeflateCompressor
SnappyCompressor
LZ4Compressor (default)
Note: If defining a table with the Encryptor encryptor, set the young generation heap (-Xmn)
parameter to a larger space to improve garbage collection. For example if running cassandrastress, set : -Xmn1600M.
The cipher_algorithm sub-option
The cipher_algorithm options and acceptable secret_key_strength for the algorithms are:
cipher_algorithm
secret_key_strength
AES/CBC/PKCS5Padding
128, 192, or 256
AES/ECB/PKCS5Padding
128, 192, or 256
DES/CBC/PKCS5Padding
56
DESede/CBC/PKCS5Padding
112 or 168
Blowfish/CBC/PKCS5Padding
32-448
RC2/CBC/PKCS5Padding
40-128
You can install custom providers for your JVM. The AES-512 is not supported out-of the box.
The key location sub-option
In DataStax Enterprise 4.0.4, the system_key option replaces the secret_key_file option and is
functionally different from the secret_key_file option. You create a file using a new dsetool command:
createsystemkey. By default, DataStax Enterprise puts the system key you create in /etc/dse/conf.
You can change the location the system key by changing the path in the dse.yaml file.
In DataStax Enterprise 4.0.3 and earlier, the secret_key_file option specified the location of the keytab
file. After specifying the data to be encrypted, a keytab file is automatically created in the directory set by
the secret_key_file. If the directory doesn’t exist, it is created. A failure to create the directory probably
indicates a permissions problem. By default, DataStax Enterprise puts the keytab file in the /etc/dse/
conf, but it can reside in any directory.
Example values in the keytab file are:
AES/ECB/PKCS5Padding:256:bxegm8vh4wE3S2hO9J36RL2gIdBLx0O46J/QmoC3W3U=
AES/CBC/PKCS5Padding:256:FUhaiy7NGB8oeSfe7cOo3hhvojVl2ijI/wbBCFH6hsE= RC2/CBC/
PKCS5Padding:128:5Iw8EW3GqE6y/6BgIc3tLw==
Deleting, moving, or changing the data in the keytab file causes errors when the node restarts and you lose
all your data. Consider storing the file on a network server or encrypting the entire file system of the nodes
using a third-party tool.
The chunk_length_kb sub-option
On disk, SSTables are encrypted and compressed by block (to allow random reads). This subproperty of
compression defines the size (in KB) of the block and is a power of 2. Values larger than the default value
might improve the compression rate, but increases the minimum size of data to be read from disk when a
read occurs. The default value (64) is a good middle-ground for compressing tables.
33
Security
Using just encryption and no compression, the size of SSTables are larger than they would be if you
combined compression. For example, on DataStax Enterprise 4.0.3 and earlier, this set of options might be
used:
Example of valid encryptions in DataStax Enterprise 4.0.3 and earlier
•
•
•
•
•
sstable_compression = EncryptingDeflateCompressor
cipher_algorithm = 'AES/CBC/PKCS5Padding'
secret_key_strength = 256
secret_key_file = '/home/automaton/newencrypt/keyfile'
chunk_length_kb = 128
In DataStax Enterprise 4.0.4 and later, omit the secret_key_file option. During creation of the table,
specifying the location of keytab file, which contains the system key, is unnecessary. DataStax Enterprise
looks for the system key as specified in dse.yaml file.
Example of valid encrpytion options in DataStax Enterprise 4.0.4 and later
•
•
•
•
sstable_compression = EncryptingDeflateCompressor
cipher_algorithm = 'AES/CBC/PKCS5Padding'
secret_key_strength = 256
chunk_length_kb = 128
The iv_length sub-option
Not all algorithms allow you to set this sub-option, and most complain if it is not set to 16 bytes. Either use
16 or accept the default.
The syntax for setting this sub-option is similar to setting a compression algorithm to compress data.
ALTER TABLE users
WITH compression =
{ 'sstable_compression' : 'EncryptingSnappyCompressor',
'cipher_algorithm' : 'AES/ECB/PKCS5Padding',
'secret_key_strength' : 128,
'iv_length' : 16 };
Using SolrJ Auth
Follow instructions in the solrj-auth-README.md file to use the SolrJ-Auth libraries to implement
encryption. The SolrJ-auth-README.md file is located in the following directories:
•
•
•
Debian installations: /usr/share/doc/dse-libsolr*
RHEL-based installations: /usr/share/doc/dse-libsolr
Binary installations: resources/solr
These SolrJ-Auth libraries are included in the DataStax Enterprise distribution:
•
•
Debian installations: /usr/share/dse/clients
Binary installations: install_location/clients
The SolrJ-Auth code is now public.
Migrating encrypted tables
Steps to migrate encrypted tables from earlier versions.
After upgrading to DataStax Enterprise 4.0.4, perform the following steps to update to the new version of
transparent data encryption, which streams encrypted data between nodes correctly.
Procedure for upgrading encrypted tables
1. Upgrade the cluster to DataStax Enterprise 4.0.4, following instructions in the "DataStax Upgrade
Guide."
34
Security
2. Restart the cluster as described in the Upgrade Guide.
3. Follow steps 4-5 in "Encrypting data in DataStax Enterprise 4.0.4 and later" to create a system key and
distribute it to all nodes in the cluster.
4. Check that the dse_system.encrypted_keys table was created as shown in step 7 in "Encrypting data in
DataStax Enterprise 4.0.4 and later" .
5. If the dse_system.encrypted_keys table was created, go to the next step; otherwise, create the table
manually.
CREATE KEYSPACE IF NOT EXISTS dse_system WITH replication = {'class':
'EverywhereStrategy'};
USE dse_system;
CREATE TABLE IF NOT EXISTS encrypted_keys (
key_file text,
cipher text,
strength int,
key_id timeuuid,
key text,
PRIMARY KEY (key_file, cipher, strength, key_id)
);
6. Rewrite all SSTables using the new version of transparent data encryption.
$ nodetool upgradesstables --include-all-sstables
If you need to restore the dse_system.encrypted_keys table, load the table. Do not truncate or delete
anything.
Configuring and using data auditing
Auditing is implemented as a log4j-based integration.
About this task
Auditing is implemented as a log4j-based integration. DataStax Enterprise places the audit log in the
directory indicated by a log4j.property. After the file reaches a threshold, it rolls over, and the file name is
changed. The file names include a numerical suffix determined by the maxBackupIndex.
The audit logger logs information on the node set up for logging. For example, node 0 has audit turned on,
node 1 does not. Issuing updates and other commands on node 1 does not generally show up on node 0’s
audit log. To get the maximum information from data auditing, turn on data auditing on every node. The
log4j supports data stored on the file system or in Cassandra.
Auditing is configured through a text file in the file system, so the file is vulnerable to OS-level security
breaches. Store the file on an OS-level encrypted file system using Vormetric, for example, to secure it.
Audit logging of queries and prepared statements submitted to the DataStax Java Driver, which uses the
CQL binary protocol, is supported.
Configuring data auditing
You can configure which categories of audit events should be logged and also whether operations against
any specific keyspaces should be omitted from audit logging.
Procedure
1. Open the log4j-server.properties file in the following directory.
•
•
Packaged installs:/etc/dse/cassandra
Tarball installs:/resources/cassandra/conf
35
Security
2. To configure data auditing, uncomment these properties, and ensure that the default properties are set.
Property
Default
Description
log4j.logger.DataAudit
INFO, A
Produce INFO-level logs.
log4j.additivity.DataAudit
false
Prevents logging to the root appender.
log4j.appender.A
org.apache.log4j.RollingFileAppender
Prevents logging to the root appender.
log4j.appender.A.File
/var/log/cassandra/audit.log
log4j.appender.A.bufferedIOtrue
Sets the file and path of the log file.
True improves performance but will not
be real time; set to false for testing.
To disable data auditing, comment out log4j.logger.DataAudit, log4j.additivity.DataAudit, and
log4jappender.A. This removes almost all auditing overhead. The Log4J audit logger logs at INFO
level, so the DataAudit logger must be configured at INFO (or lower) level in log4j-server.properties.
Setting the logger to a higher level, such as WARN, prevents any log events from being recorded, but
it does not completely disable the data auditing. Some overhead occurs beyond that caused by regular
processing.
3. Set other general options to tune the logging, for example uncomment these properties and accept the
following defaults:
• log4j.appender.A.maxFileSize=200MB
• log4j.appender.A.maxBackupIndex=5
• log4j.appender.A.layout=org.apache.log4j.PatternLayout
• log4j.appender.A.layout.ConversionPattern=%m%n
• log4j.appender.A.filter.1=com.datastax.bdp.cassandra.audit.AuditLogFilter
4. Uncomment and set log4j.appender.A.filter.1.ActiveCategories to ALL or to a combination of these
settings:
Setting
Logging
ADMIN
Logs describe schema versions, cluster name, version, ring, and other admin
events
ALL
Logs everything: DDL, DML, queries, and errors
AUTH
Logs login events
DML
Logs insert, update, delete and other DML events
DDL
Logs object and user create, alter, drop, and other DDL events
DCL
Logs grant, revoke, create user, drop user, and list users events
QUERY
Logs all queries
Set the ActiveCategories property to a comma separated list of the categories to include in the audit log
output. By default, this list is empty so unless specified, no events are included in the log. Events are
generated even if not included in the log, so set this property.
5. You can disable logging for specific keyspaces. Set this property as follows to prevent logging to
specified keyspaces:
log4j.appender.A.filter.1.ExemptKeyspaces=do_not_log,also_do_not_log
To prevent the audit logger from logging information about itself when using the Cassandra log4j
appender, exempt the keyspace from the appender logs.
Results
The audit log section of the log4j-server.properties file should look something like this:
36
Security
log4j.logger.DataAudit=INFO, A
log4j.additivity.DataAudit=false
log4j.appender.A=org.apache.log4j.RollingFileAppender
log4j.appender.A.File=/var/log/cassandra/audit.log
log4j.appender.A.bufferedIO=true
log4j.appender.A.maxFileSize=200MB
log4j.appender.A.maxBackupIndex=5
log4j.appender.A.layout=org.apache.log4j.PatternLayout
log4j.appender.A.layout.ConversionPattern=%m%n
log4j.appender.A.filter.1=com.datastax.bdp.cassandra.audit.AuditLogFilter
log4j.appender.A.filter.1.ActiveCategories=ALL
log4j.appender.A.filter.1.ExemptKeyspaces=do_not_log,also_do_not_log
Restart the node to see changes in the log.
Formats of logs
A description of log formats, examples, and batch updates.
The log format is a simple set of pipe-delimited name/value pairs. The pairs themselves are separated by
the pipe symbol ("|"), and the name and value portions of each pair are separated by a colon. A name/
value pair, or field, is only included in the log line if a value exists for that particular event. Some fields
always have a value, and are always present. Others might not be relevant for a given operation. The order
in which fields appear (when present) in the log line is predictable to make parsing with automated tools
easier. For example, the text of CQL statements is unquoted but if present, is always the last field in the log
line.
Field Label
Field Value
Optional
host
dse node address
no
source
client address
no
user
authenticated user
no
timestamp
system time of log event
no
category
DML/DDL/QUERY for
example
no
type
API level operation
no
batch
batch id
yes
ks
keyspace
yes
cf
column family
yes
operation
textual description
yes
The textual description value for the operation field label is currently only present for CQL.
Auditing is completely separate from authorization, although the data points logged include the
client address and authenticated user, which may be a generic user if the default authenticator is not
overridden. Logging of requests can be activated for any or all of the first list of categories covered by
log4j.appender.A.filter.1.ActiveCategories (shown in step 4 in Configuring data auditing).
CQL logging examples
Generally, SELECT queries are placed into the QUERY category. The INSERT, UPDATE, and DELETE
statements are categorized as DML. CQL statements that affect schema, such as CREATE KEYSPACE
and DROP KEYSPACE are categorized as DDL.
CQL USE
37
Security
USE dsp904;
host:/192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351003707937|category:DML|type:SET_KS|ks:dsp904|operation:use
dsp904;
CLI USE
USE dsp904;
host:/192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351004648848|category:DML|type:SET_KS|ks:dsp904
CQL query
SELECT * FROM t0;
host:/192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351003741953|category:QUERY|type:CQL_SELECT|ks:dsp904|cf:t0|
operation:select * from t0;
CQL BATCH
BEGIN BATCH
INSERT INTO t0(id, field0) VALUES (0, 'foo')
INSERT INTO t0(id, field0) VALUES (1, 'bar')
DELETE FROM t1 WHERE id = 2
APPLY BATCH;
host:192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351005482412|category:DML|type:CQL_UPDATE
|batch:fc386364-245a-44c0-a5ab-12f165374a89|ks:dsp904|cf:t0
|operation:INSERT INTO t0 ( id , field0 ) VALUES ( 0 , 'foo' )
host:192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351005482413|category:DML|type:CQL_UPDATE
|batch:fc386364-245a-44c0-a5ab-12f165374a89|ks:dsp904|cf:t0
|operation:INSERT INTO t0 ( id , field0 ) VALUES ( 1 , 'bar' )
host:192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351005482413|category:DML|type:CQL_DELETE
|batch:fc386364-245a-44c0-a5ab-12f165374a89|ks:dsp904|cf:t1
|operation:DELETE FROM t1 WHERE id = 2
CQL DROP KEYSPACE
DROP KEYSPACE dsp904;
host:/192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351004777354|category:DDL|type:DROP_KS
|ks:dsp904|operation:drop keyspace dsp904;
CQL prepared statement
host:/10.112.75.154|source:/127.0.0.1|user:allow_all
|timestamp:1356046999323|category:DML|type:CQL_UPDATE
|ks:ks|cf:cf|operation:INSERT INTO cf (id, name) VALUES (?, ?)
[id=1,name=vic]
Thrift batch_mutate
host:/192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351005073561|category:DML|type:INSERT
|batch:7d13a423-4c68-4238-af06-a779697088a9|ks:Keyspace1|cf:Standard1
host:/192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351005073562|category:DML|type:INSERT
|batch:7d13a423-4c68-4238-af06-a779697088a9|ks:Keyspace1|cf:Standard1
38
Security
host:/192.168.56.1|source:/192.168.56.101|user:#User allow_all groups=[]
|timestamp:1351005073562|category:DML|type:INSERT
|batch:7d13a423-4c68-4238-af06-a779697088a9|ks:Keyspace1|cf:Standard1
DataStax Java Driver queries
host:ip-10-85-22-245.ec2.internal/10.85.22.245|source:/127.0.0.1|
user:anonymous
|timestamp:1370537557052|category:DDL|type:ADD_KS
|ks:test|operation:create keyspace test with replication =
{'class':'NetworkTopologyStrategy', 'Analytics': 1};
host:ip-10-85-22-245.ec2.internal/10.85.22.245|source:/127.0.0.1|
user:anonymous
|timestamp:1370537557208|category:DDL|type:ADD_CF
|ks:test|cf:new_cf|operation:create COLUMNFAMILY test.new_cf ( id text
PRIMARY KEY , col1 int, col2 ascii, col3 int);
host:ip-10-85-22-245.ec2.internal/10.85.22.245|source:/127.0.0.1|
user:anonymous
|timestamp:1370537557236|category:DML|type:CQL_UPDATE
|ks:test|cf:new_cf|operation:insert into test.new_cf (id, col1, col2, col3)
values ('test1', 42, 'blah', 3);
host:ip-10-85-22-245.ec2.internal/10.85.22.245|source:/127.0.0.1|
user:anonymous
|timestamp:1370537704885|category:QUERY|type:CQL_SELECT
|ks:test|cf:new_cf|operation:select * from test.new_cf;
Batch updates
Batch updates, whether received via a Thrift batch_mutate call, or in CQL BEGIN BATCH....APPLY
BATCH block, are logged in the following way: A UUID is generated for the batch, then each individual
operation is reported separately, with an extra field containing the batch id.
Configuring auditing for a DSE Search/Solr cluster
The filter-mapping element in the Solr web.xml file enables auditing.
About this task
By default, DSE Search/Solr nodes need no configuration for data auditing except setting up the log4jserver.properties file. If the filter-mapping element in the Solr web.xml file is commented out, the auditor
cannot log anything from Solr and you need to configure auditing as described in the next section.
Procedure
If necessary, uncomment the filter-mapping element in the Solr web.xml.
<filter-mapping>
<filter-name>DseAuditLoggingFilter</filter-name>
<url-pattern>/*</url-pattern>
</filter-mapping>
The Solr web.xml is located in the following directory:
•
•
Packaged installs:/usr/share/dse/solr/web/solr/WEB-INF/web.xml
Tarball installs:/resources/solr/web/solr/WEB-INF/web.xml
Solr Audit Log example
Here is an example of the data audit log of a Solr query:
host:/10.245.214.159|source:127.0.0.1|user:jdoe|timestamp:1356045339910|
category:QUERY
39
Security
|type:SOLR_QUERY|ks:wiki|cf:solr|operation:/wiki.solr/select/?
q=body:trains
Configuring and using internal authentication
Internal authentication is based on Cassandra-controlled login accounts and passwords.
Like object permission management that uses internal authorization, internal authentication is based on
Cassandra-controlled login accounts and passwords. Internal authentication is supported on the following
clients when you provide a user name and password to start up the client:
•
•
•
•
•
•
Astyanax
cassandra-cli
cqlsh
DataStax Java and C# drivers
Hector
pycassa
Internal authentication stores user names and bcrypt-hashed passwords in the system_auth.credentials
table.
Limitations
DataStax Enterprise 4.0 - 4.0.2 does not support internal authentication in Solr, the dsetool, and Hadoop
utilities. DataStax Enterprise 4.0.3 and later provides internal authentication support for some Hadoop
tools.
Authentication for Hadoop tools
After configuring authentication in DataStax Enterprise 4.0.3, starting Hadoop requires a user name and
password. These login credentials can be provided using a file or a command line option as follows:
Using a file
You can provide the user name and password by creating a file named .dserc in your DataStax
Enterprise home directory. The ~/.dserc file contains the user name and password:
username=<username>
password=<password>
Using the command line
If a ~/.dserc file does not exist, use these options on the dse command line to provide the login credentials
as follows:
dse hadoop <command> -Dcassandra.username=<username> Dcassandra.password=<password> <other options>
dse hive <hive options> -hiveconf cassandra.username=<username> -hiveconf
cassandra.password=<password>
dse pig -Dcassandra.username=<username> -Dcassandra.password=<password> <pig
options>
dse sqoop <sqoop options> --cassandra-username=<username> --cassandrapassword=<password>
The dse command reference covers other options.
40
Security
Hadoop tool authentication limitations
•
•
•
Internal authentication is not supported for Mahout.
Using internal authentication to run the hadoop jar command is not supported.
The hadoop jar command accepts only the jar file name as an option, and rejects other options
such as username and password. The main class in the jar is responsible for making sure that the
credentials are applied to the job configuration.
In Pig scripts that use the custom storage handlers CqlStorage and CassandraStorage storage
handlers, provide credentials in the URL of the URL-encoded prepared statement:
cql://<username>:<password>@<keyspace>/<columnfamily>
cassandra://<username>:<password>@<keyspace>/<columnfamily>
Use this method of providing authentication for Pig commands regardless of the mechanism you use for
passing credentials to Pig.
Configuring internal authentication and authorization
Steps to configure internal authentication and authorization.
About this task
You must set internal authentication and authorization at the same time. After setting the Authorizer and
the Authenticator in the cassandra.yaml file, you can set object permissions, as described in Managing
object permissions using internal authorization.
Procedure
Perform the first three steps on every node.
1. Change the authenticator option in the cassandra.yaml to the native Cassandra
PasswordAuthenticator by uncommenting only the PasswordAuthenticator:
authenticator: org.apache.cassandra.auth.PasswordAuthenticator
You can use any authenticator except AllowAll.
2. Change the authorizer option by commenting the AllowAllAuthorizer and adding the
CassandraAuthorizer:
#authorizer: org.apache.cassandra.auth.AllowAllAuthorizer
authorizer: org.apache.cassandra.auth.CassandraAuthorizer
3. Restart the node.
Note: You can enable internal authorization on existing clusters with no downtime.
4. One one node, configure the system_auth keyspace replication factor.
Fetching permissions can be an expensive operation. If necessary, adjust the validity period for
permissions caching by setting the permissions_validity_in_ms option in the cassandra.yaml. You
can also disable permission caching by setting this option to 0.
5. Run a full repair of the system_auth keyspace.
6. Start cqlsh using the same superuser name and password (cassandra) that you use to start the
supported client. For example, to start cqlsh on Linux:
./cqlsh -u cassandra -p cassandra
7. Change the superuser's user name and password.
Changing the default superuser
You can change the default superuser from the default cassandra user.
41
Security
About this task
By default, each installation of Cassandra includes a superuser account named cassandra whose
password is also cassandra. A superuser grants initial permissions to access Cassandra data, and
subsequently a user may or may not be given the permission to grant/revoke permissions.
Procedure
1.
2.
3.
4.
Configure internal authentication if you have not already done so.
Create another superuser, not named cassandra, using the CREATE USER command.
Log in as that new superuser.
Change the cassandra user password to something long and incomprehensible, and then forget about
it. It won't be used again.
5. Take away the cassandra user's superuser status.
6. Now, that the superuser password is secure, set up user accounts and authorize users to access the
database objects by using CQL to grant them permissions on those objects.
CQL supports the following authentication statements:
•
•
•
•
alter-user
create-user
drop-user
list-users
Enable internal security without downtime
TransitionalAuthenticator and TransitionalAuthorizer allow internal authentication and authorization to be
enabled without downtime or modification to client code or configuration.
About this task
The TransitionalAuthenticator and TransitionalAuthorizer allow internal authentication and
authorization to be enabled without downtime or modification to client code or configuration.
Procedure
1. On each node, in the cassandra.yaml file:
•
Set the authenticator to
com.datastax.bdp.cassandra.auth.TransitionalAuthenticator.
• Set the authorizer to com.datastax.bdp.cassandra.auth.TransitionalAuthorizer.
2. Perform a rolling restart.
3. Run a full repair of the system_auth keyspace
4. Once the restarts are complete, use cqlsh with the default superuser login to setup the users,
credentials, and permissions.
5. Once the setup is complete, edit the cassandra.yaml file again and perform another rolling restart:
• Change the authenticator to org.apache.cassandra.auth.PasswordAuthenticator.
• Change the authorizer to org.apache.cassandra.auth.CassandraAuthorizer.
6. After the restarts have completed, remove the default superuser and create at least one new superuser.
Logging in with cqlsh
Create a cqlshrc file to pass default login information.
42
Security
About this task
To avoid having to pass credentials for every login using cqlsh, you can create a cqlshrc file your
~/.cassandra directory. When present, it passes default login information to cqlsh. For example:
Procedure
Create the cqlshrc file with the following in formation:
[authentication]
username = username
password = password
Be sure to set the correct permissions and secure this file so that no unauthorized users can gain
access to database login information.
Note: Sample cqlshrc files are available in the following directories:
•
•
Packaged installs: /usr/share/doc/dse-libcassandra
Tarball installs: install_location/resources/cassandra/conf
Managing object permissions using internal authorization
Use GRANT/REVOKE to grant or revoke permissions to access Cassandra data.
About this task
You use the familiar relational database GRANT/REVOKE paradigm to grant or revoke permissions to
access Cassandra data. A superuser grants initial permissions, and subsequently a user may or may not
be given the permission to grant/revoke permissions. Object permission management is independent of
authentication (works with Kerberos or Cassandra).
CQL supports the following authorization statements:
•
•
•
GRANT
LIST PERMISSIONS
REVOKE
Accessing system resources
Read access to these system tables is implicitly given to every authenticated user because the tables are
used by most Cassandra tools:
•
•
•
•
•
system.schema_keyspace
system.schema_columns
system.schema_columnfamilies
system.local
system.peers
Configuration
CassandraAuthorizer is one of many possible IAuthorizer implementations, and the one that stores
permissions in the system_auth.permissions table to support all authorization-related CQL statements.
Configuration consists mainly of changing the authorizer option in the cassandra.yaml as described in
Configuring internal authentication and authorization.
Note: You must set internal authentication and authorization at the same time.
43
Security
Configuring system_auth keyspace replication
The system_auth and dse_security keyspaces store security authentication and authorization information.
About this task
Cassandra uses the system_auth keyspace for storing security authentication and authorization
information. If you use the following authenticator/authorizer, you must set the replication factor with a
keyspace command such as ALTER KEYSPACE to prevent a potential problem logging into a secure
cluster:
•
•
authenticator: org.apache.cassandra.auth.PasswordAuthenticator: the users' hashed
passwords in system_auth.credentials table
authorizer: org.apache.cassandra.auth.CassandraAuthorizer: the users' permissions in
system_auth.permissions table
Setting the replication factor
About this task
Do not use the default replication factor of 1 for the system_auth keyspace. In a multi-node cluster, using
the default of 1 precludes logging into any node when the node that stores the user data is down. For most
system_auth queries, Cassandra uses a consistency level of LOCAL_ONE and uses QUORUM for the
default cassandrasuperuser; see Configuring data consistency.
Procedure
Set the replication factor based on one of the following examples depending on your environment:
•
SimpleStrategy example:
•
ALTER KEYSPACE "system_auth"
WITH REPLICATION = { 'class' : 'SimpleStrategy', 'replication_factor' :
3 };
NetworkTopologyStrategy example:
ALTER KEYSPACE "system_auth"
WITH REPLICATION = {'class' : 'NetworkTopologyStrategy', 'dc1' : 3,
'dc2' : 2};
Configuring firewall port access
Opening the required ports to allow communication between the nodes.
About this task
If you have a firewall running on the nodes in your Cassandra or DataStax Enterprise cluster, you must
open up the following ports to allow communication between the nodes, including certain Cassandra ports.
If this isn't done, when you start Cassandra (or Hadoop in DataStax Enterprise) on a node, the node will
act as a standalone database server rather than joining the database cluster.
Procedure
Open the following ports:
Port Description
Public Facing Ports
44
Configurable in
Security
Port Description
Configurable in
22
See your OS
documentation on
sshd.
SSH (default)
DataStax Enterprise public ports
8012 Hadoop Job Tracker client port. The Job Tracker listens on this port for cassandra.yaml
job submissions and communications from task trackers; allows traffic
See Setting the job
from each Analytics node in a cluster.
tracker node.
8983 Solr port and Demo applications website port (Portfolio, Search,
Search log)
50030Hadoop Job Tracker website port. The Job Tracker listens on this port
for HTTP requests. If initiated from the OpsCenter, these requests are
proxied through the opscenterd daemon; otherwise, they come directly
from the browser. [1]
mapredsite.xml using the
mapred.job.tracker.http.address
property.
50060Hadoop Task Tracker website port. Each Task Tracker listens on
this port for HTTP requests coming directly from the browser and not
proxied by the opscenterd daemon. [1]
mapredsite.xml using the
mapred.task.tracker.http.address
property.
OpsCenter public ports
8888 OpsCenter website. The opscenterd daemon listens on this port for
HTTP requests coming directly from the browser. [1]
opscenterd.conf
Inter-node Ports
Cassandra inter-node ports
1024 JMX reconnection/loopback ports. Please read the description for port
7199.
65355
7000 Cassandra inter-node cluster communication.
cassandra.yaml
See storage_port.
7001 Cassandra SSL inter-node cluster communication.
cassandra.yaml
See ssl_storage_port.
7199 Cassandra JMX monitoring port. After the initial handshake, the JMX
protocol requires that the client reconnects on a randomly chosen port
(1024+). Open this port only if you want to remotely connect to the
node via JMX. Running nodetool or opscenter agent locally does not
require these ports to be open.
cassandra-env.sh
See JMX options in
Tuning Java resources.
Note: Starting with Java 7u4, you can specify the port used by
JMX rather than a randomly assigned port. The standard RMI
(Remote Method Invocation) registry port for JMX is set by the
com.sun.management.jmxremote.port property. Use the
com.sun.management.jmxremote.rmi.port property to
specify the port used by JMX.
45
Security
Port Description
Configurable in
9160 Cassandra client port (Thrift). OpsCenter agents makes Thrift requests cassandra.yaml
to their local node on this port. Additionally, the port can be used by
See rpc_port.
the opscenterd daemon to make Thrift requests to each node in the
cluster.
DataStax Enterprise inter-node ports
8984 Netty server port.
dse.yaml See Shard
transport options for
DSE Search/Solr
communications.
9042 CQL native clients port.
cassandra.yaml
See
native_transport_port.
9290 Hadoop Job Tracker Thrift port. The Job Tracker listens on this port for
Thrift requests coming from the opscenterd daemon.
10000Hive server port.
OpsCenter specific inter-node
50031OpsCenter HTTP proxy for Job Tracker. The opscenterd daemon
listens on this port for incoming HTTP requests from the browser when
viewing the Hadoop Job Tracker page directly. [1]
61620OpsCenter monitoring port. The opscenterd daemon listens on this
port for TCP traffic coming from the agent. [1]
61621OpsCenter agent port. The agents listen on this port for SSL traffic
initiated by OpsCenter. [1]
[1] See OpsCenter and DataStax agent ports.
46
Set with the -p
option in the dse
hive --service
hiveserver
command or configure
in hive-site.xml.
DSE Analytics with Hadoop
DSE Analytics with Hadoop
DSE Hadoop topics.
Getting started with Analytics and Hadoop in DataStax Enterprise
The Hadoop component in DataStax Enterprise enables analytics to be run across DataStax Enterprise's
distributed, shared-nothing architecture. Instead of using the Hadoop Distributed File System (HDFS),
DataStax Enterprise uses Cassandra File System (CFS) keyspaces for the underlying storage layer.
About this task
You can run analytics on your Cassandra data using Hadoop, which is integrated into DataStax Enterprise.
The Hadoop component in DataStax Enterprise is not meant to be a full Hadoop distribution, but rather
enables analytics to be run across DataStax Enterprise's distributed, shared-nothing architecture. Instead
of using the Hadoop Distributed File System (HDFS), DataStax Enterprise uses Cassandra File System
(CassandraFS) keyspaces for the underlying storage layer. This provides replication, data location
awareness, and takes full advantage of Cassandra's peer-to-peer architecture.
DataStax Enterprise 4.0.3 and later supports internal authentication for running hadoop, hive, pig, and
sqoop commands.
DataStax Enterprise supports running analytics on Cassandra data with the following Hadoop components:
•
•
•
•
MapReduce
Hive for running HiveQL queries on Cassandra data
Pig for exploring very large data sets
Apache Mahout for machine learning applications
To get started using DSE Analytics with Hadoop, run the Portfolio Manager demo.
DataStax Enterprise 4.0 turns off virtual nodes (vnodes) by default. DataStax does not recommend
turning on vnodes for Hadoop or Solr nodes, but you can use vnodes for any Cassandra-only cluster, or
a Cassandra-only data center in a mixed Hadoop/Solr/Cassandra deployment. To run Hadoop, disable
virtual nodes.
Performance enhancement
Performance reading MapReduce files in the CassandraFS has been improved in DataStax Enterprise 4.0.
These files are now stored in the page cache, making the files available on the next read.
Starting a DSE Analytics node
The way you start up a DSE Analytics node depends on the type of installation:
•
Tarball installs:
From the installation directory:
•
$ bin/dse cassandra -t
Packaged installation:
1. Enable Hadoop mode by setting this option in /etc/default/dse:
HADOOP_ENABLED=1
2. Use this command to start the service:
$ sudo service dse start
47
DSE Analytics with Hadoop
Stopping a DSE Analytics node
The way you stop up a DSE Analytics node depends on the type of installation:
•
Tarball installs:
1. From the install directory:
$ bin/dse cassandra-stop
2. Check that the dse process has stopped.
$ ps auwx | grep dse
If the dse process stopped, the output should be minimal, for example:
jdoe
12390 0.0 0.0
2432768
620 s000
R+ 2:17PM
0:00.00 grep dse
If the output indicates that the dse process is not stopped, rerun the cassandra-stop command using
the process ID (PID) from the top of the output.
•
bin/dse cassandra-stop PID
Packaged installation:
$ sudo service dse stop
Hadoop getting started tutorial
A tutorial about using Apache Hadoop embedded in DataStax Enterprise.
About this task
In this tutorial, you download a text file containing a State of the Union speech and run a classic
MapReduce job that counts the words in the file and creates a sorted list of word/count pairs as output. The
mapper and reducer are provided in a JAR file. Download the State of the Union speech now.
This tutorial assumes you started an analytics node on Linux. Also, the tutorial assumes you have
permission to perform Hadoop and other DataStax Enterprise operations, for example, or that you preface
commands with sudo if necessary.
Procedure
1. Unzip the downloaded obama.txt.zip file into a directory of your choice on your file system.
This file will be the input for the MapReduce job.
2. Create a directory in the CassandraFS for the input file using the dse command version of the familiar
hadoop fs command. On a Linux tarball installation, for example:
$ cd install_location
$ bin/dse hadoop fs -mkdir /user/hadoop/wordcount/input
3. Copy the input file that you downloaded to the CassandraFS.
$ bin/dse hadoop fs -copyFromLocal
path/obama.txt
/user/hadoop/wordcount/input
4. Check the version number of the hadoop-examples-version.jar. On tarball installations, the JAR
is in the DataStax Enterprise resources directory. On packaged and AMI installations, the JAR is in the
/usr/share/dse/hadoop/lib directory.
5. Get usage information about how to run the MapReduce job from the jar file.
$ bin/dse hadoop jar /install_location/resources/hadoop/hadoopexamples-1.0.4.9.jar wordcount
The output is:
2013-10-02 12:40:16.983 java[9505:1703] Unable to load realm info from
SCDynamicStore
48
DSE Analytics with Hadoop
Usage: wordcount <in> <out>
If you see the SCDynamic Store message, just ignore it. The internet provides information about the
message.
6. Run the Hadoop word count example in the JAR.
$ bin/dse hadoop jar
/install_location/resources/hadoop/hadoop-examples-1.0.4.9.jar wordcount
/user/hadoop/wordcount/input
/user/hadoop/wordcount/output
The output is:
13/10/02 12:40:36 INFO input.FileInputFormat: Total input paths to process :
0
13/10/02 12:40:36 INFO mapred.JobClient: Running job: job_201310020848_0002
13/10/02 12:40:37 INFO mapred.JobClient: map 0% reduce 0%
. . .
13/10/02 12:40:55 INFO mapred.JobClient:
FILE_BYTES_WRITTEN=19164
13/10/02 12:40:55 INFO mapred.JobClient:
Map-Reduce Framework
7. List the contents of the output directory on the CassandraFS.
$ bin/dse hadoop fs -ls /user/hadoop/wordcount/output
The output looks something like this:
Found 3 items
-rwxrwxrwx
1 root wheel
0 2013-10-02 12:58 /user/hadoop/wordcount/
output/_SUCCESS
drwxrwxrwx
- root wheel
0 2013-10-02 12:57 /user/hadoop/wordcount/
output/_logs
-rwxrwxrwx
1 root wheel 24528 2013-10-02 12:58 /user/hadoop/wordcount/
output/part-r-00000
8. Using the output file name from the directory listing, get more information using the dsetool utility.
$ bin/dsetool checkcfs /user/hadoop/wordcount/output/part-r-00000
The output is:
Path: cfs://127.0.0.1/user/hadoop/wordcount/output/part-r-00000
INode header:
File type: FILE
User: root
Group: wheel
Permissions: rwxrwxrwx (777)
Block size: 67108864
Compressed: true
First save: true
Modification time: Wed Mar 02 12:58:05 PDT 2014
INode:
Block count: 1
Blocks:
subblocks
length
start
end
(B) f2fa9d90-2b9c-11e3-9ccb-73ded3cb6170:
1
24528
0
24528
f3030200-2b9c-11e3-9ccb-73ded3cb6170:
24528
0
24528
Block locations:
f2fa9d90-2b9c-11e3-9ccb-73ded3cb6170: [localhost]
Data:
All data blocks ok.
9. Finally, look at the output of the MapReduce job--the list of word/count pairs using a familiar Hadoop
command.
$ bin/dse hadoop fs -cat /user/hadoop/wordcount/output/part-r-00000
The output is:
"D." 1
"Don't 1
49
DSE Analytics with Hadoop
"I 4
. . .
Analytics node configuration
Steps to configure analytic Hadoop nodes.
About this task
Important configuration changes, excluding those related to the job tracker, are:
•
•
•
•
Disabling virtual nodes
Setting the replication factor
Configuring the verbosity of log messages
Connecting to non-standard Cassandra native port
Advanced users can also configure DataStax Enterprise to run jobs remotely.
DataStax Enterprise turns off virtual nodes (vnodes) by default. DataStax does not recommend turning on
vnodes for Hadoop or Solr nodes, but you can use vnodes for any Cassandra-only cluster, or a Cassandraonly data center in a mixed Hadoop/Solr/Cassandra deployment. If you have enabled virtual nodes on
Hadoop nodes, disable virtual nodes before using the cluster.
Setting the replication factor
The default replication for the HiveMetaStore, cfs, and cfs_archive system keyspaces is 1. A replication
factor of 1 using the default data center Analytics is configured for development and testing of a single
node, not for a production environment. For production clusters, increase the replication factor to at least 2.
The higher replication factor ensures resilience to single-node failures. For example:
ALTER KEYSPACE cfs
WITH REPLICATION = {'class' : 'NetworkTopologyStrategy', 'dc1' : 3};
Configuring the verbosity of log messages
To adjust the verbosity of log messages for Hadoop map/reduce tasks, add the following settings to the
log4j.properties file on each analytic node:
log4j.logger.org.apache.hadoop.mapred=WARN
log4j.logger.org.apache.hadoop.filecache=WARN
Default Hadoop log4j-server.properties locations:
•
•
Installer-Services and Package installations: /etc/dse/hadoop/
Installer-No Services and Tarball installations: install_location/resources/hadoop/conf/
Connecting to non-standard Cassandra native port
If the Cassandra native port was changed to a port other than the default port 9042, you must change the
cassandra.input.native.port configuration setting for Hive and Hadoop to use the non-default port. The
following examples change the Cassandra native port protocol connections to use port 9999.
•
Inside the Hive shell, set the port after starting DSE Hive:
•
$ dse hive
hive> set cassandra.input.native.port=9999;
General Hive, add cassandra.input.native.port to the hive-site.xml file:
<property>
<name>cassandra.input.native.port</name>
<value>9999</value>
</property>
50
DSE Analytics with Hadoop
•
For Hadoop, add cassandra.input.native.port to the core-site.xml file:
<property>
<name>cassandra.input.native.port</name>
<value>9999</value>
</property>
Configuration for running jobs on a remote cluster
This information is intended for advanced users.
Procedure
To connect to external addresses:
1. Make sure that the hostname resolution works properly on the localhost for the remote cluster nodes.
2. Copy the dse-core-default.xml and dse-mapred-default.xml files from any working remote
cluster node to your local Hadoop conf directory.
3. Run the job using dse hadoop.
4. If you need to override the JT location or if DataStax Enterprise cannot automatically detect the JT
location, before running the job, define the HADOOP_JT environment variable:
$ export HADOOP_JT=jobtracker host:jobtracker port dse hadoop jar ....
5. If you need to connect to many different remote clusters from the same host:
a) Before starting the job, copy the remote Hadoop conf directories fully to the local node (into different
locations).
b) Select the appropriate location by defining HADOOP_CONF_DIR.
Changing the Hadoop log directory
Add the HADOOP_LOG_DIR environment variable to the dse-env.sh file to recognize changes to the
default log directory used by the Hadoop component that is integrated into DataStax Enterprise.
About this task
DataStax Enterprise does not recognize changes to the default log directory used by the Hadoop
component integrated into DataStax Enterprise 4.0 unless you add the HADOOP_LOG_DIR environment
variable to the dse-env.sh file, as described in this topic.
The Hadoop environment variables are set in the hadoop-env.sh file. In this file, the
HADOOP_LOG_DIR default configuration is:
•
•
Tarball installs: /var/log/hadoop
Packaged installs: /etc/dse/hadoop/hadoop-env.sh
If you change the default Hadoop log directory environment variable in hadoop-env.sh and restart
DataStax Enterprise, the change will not be recognized.
The workaround, if you want to change the default Hadoop log directory, is to add HADOOP_LOG_DIR to
the following file:
•
•
Tarball installs: install_dir/bin/dse-env.sh
Packaged installs: /etc/dse/dse-env.sh
Procedure
1. In the dse-env.sh file, you will see comments about exactly where to add the command to configure
the environment variable. For example:
#!/bin/sh
51
DSE Analytics with Hadoop
# Add any environment overrides you need here. This is where users
# may set third-party variables such as HADOOP_LOG_DIR
export HADOOP_LOG_DIR=/var/log/hadoop/new_log_location
# ==================================
# don't change after this.
if [ -r "`dirname "$0"`/dse.in.sh" ]; then
. . .
2. Restart DataStax Enterprise after configuring the new log location.
In a packaged installation, DataStax Enterprise loads the environment variable change using /usr/
share/dse/dse.in.sh after you restart the node.
Portfolio Manager demo
A demonstration of a financial application where users can actively create and manage a portfolio of
stocks.
About this task
The use case is a financial application where users can actively create and manage a portfolio of stocks.
On the Cassandra OLTP (online transaction processing) side, each portfolio contains a list of stocks, the
number of shares purchased, and the purchase price. The demo's pricer utility simulates real-time stock
data where each portfolio updates based on its overall value and the percentage of gain or loss compared
to the purchase price. This utility also generates 100 days of historical market data (the end-of-day price)
for each stock. On the DSE OLAP (online analytical processing) side, a Hive MapReduce job calculates
the greatest historical 10 day loss period for each portfolio, which is an indicator of the risk associated
with a portfolio. This information is then fed back into the real-time application to allow customers to better
gauge their potential losses.
Before you begin
Before running the demo, make sure the following items have been completed:
•
•
A single-node or multiple node instance of DataStax Enterprise is installed.
The cluster is configured and running.
Procedure
To run the demo:
1. Check which mode DataStax Enterprise is running in:
•
Packaged installs:
•
$ nodetool status
Tarball installs:
$ bin/nodetool status
2. If the data center is not Analytics, stop the node:
•
Packaged installs:
•
$ sudo service dse stop
Tarball installs: From the install location:
$ install_location/bin/dse cassandra-stop
52
## Use sudo if necessary
DSE Analytics with Hadoop
In the unlikely event that the cassandra-stop command fails because it cannot find the process
DataStax Enterprise Java process ID (PID), the output instructs you to find the DataStax Enterprise
Java process ID (PID) manually, and stop the process using its PID number.
$ ps auwx | grep dse
$ bin/dse cassandra-stop -p PID
3. Start DataStax Enterprise as an analytics node:
•
Packaged installs: Edit the /etc/default/dse file, and change
HADOOP_ENABLED=0
to
HADOOP_ENABLED=1
and then start DataStax Enterprise:
•
$ sudo service dse start
Tarball installs:
$ install_location/bin/dse cassandra -t
4. Go to the portfolio manager demo directory:
a) Test whether the dse-demos directory exists.
$ file ~/dse-demos
b) If the directory does not exist:
$ install_dir/bin/make-dse-demos.sh
The dse-demos directory is created in your current directory.
•
•
GUI/Text Services and package installations: $ cd ~/dse-demos/portfolio_manager
GUI/Text No Services and tarball installations: $ cd install_location/dse-demos/
portfolio_manager
• Ran script to create directory $ cd dir-you-ran-script-in/dse-demos
5. Start the web service:
$ cd website
$ ./start
6. Open a browser and go to http://localhost:8983/portfolio.
The real-time Portfolio Manager demo application is displayed.
7. Open another terminal.
8. Start Hive and run the MapReduce job for the demo in Hive.
53
DSE Analytics with Hadoop
•
•
GUI/Text Services and package installations: $ dse hive -f /usr/share/dse-demos/
portfolio_manager/10_day_loss.q
GUI/Text No Services and tarball installations: $ install_location/bin/dse hive -f
install_location/demos/portfolio_manager/10_day_loss.q
The MapReduce job takes several minutes to run.
9. To watch the progress in the job tracker, open the following URL in a browser.
http://localhost:50030/jobtracker.jsp
10.After the job completes, refresh the Portfolio Manager web page.
The results of the Largest Historical 10 day Loss for each portfolio are displayed.
Using the job tracker node
DataStax Enterprise schedules a series of tasks on the analytics nodes for each MapReduce job that is
submitted to the job tracker.
About this task
For each MapReduce job submitted to the job tracker, DataStax Enterprise schedules a series of tasks on
the analytics nodes. One task tracker service per node handles the map and reduce tasks scheduled for
that node. Within a data center, the job tracker monitors the execution and status of distributed tasks that
comprise a MapReduce job.
Using multiple job tracker services
You can use multiple job tracker nodes in a cluster, one per data center. In deployments having multiple
data centers far away from each other, using multiple job trackers and multiple file systems can improve
performance by taking advantage of data locality on each cluster.
Tasks related to the job tracker are:
•
•
•
Setting the job tracker node
Managing the job tracker using dsetool commands
Changing the job tracker client port
Setting the job tracker node
Different ways to set the job tracker node.
There are several ways to set the job tracker node.
•
•
You can configure the Cassandra seeds list in the cassandra.yaml. From the IP addresses in the
seeds list of the cassandra.yaml file, DataStax Enterprise nominates the first analytics node in the
list in each data center to be the job tracker when you start the analytics node.
You can start up an analytics node using the -j option on a tarball installation. This option designates
the node being started as the job tracker node.
$ install_location/bin/dse cassandra -t -j
54
DSE Analytics with Hadoop
•
You can also use this method on a packaged installation to designate the job tracker when starting the
analytics node as a standalone process instead of a service.
You can use the dsetool movejt command.
About the reserve job tracker
DataStax Enterprise nominates a node in the cluster as a reserve job tracker for a data center. The reserve
job tracker becomes the job tracker when, for some reason, there is no local node in the data center that
can function as job tracker.
Using common hadoop commands
Common hadoop commands perform functions in the Cassandra File System (CFS) that correspond to
open source, HDFS file system shell commands.
Use common hadoop commands to perform functions in the CassandraFS that correspond to open source,
HDFS file system shell commands. The format of the URI for the CassandraFS is:
[cfs-name:][//[host]] path
•
•
•
If cfs-name is missing, cfs, which means to access the CassandraFS, is used.
If host is missing, the address of the local node is used.
If host is given, the path must start with /
For example, the following paths point to the same path in the CassandraFS:
/tmp
///tmp
cfs:/tmp
cfs:///tmp
cfs://localhost/tmp
//localhost/tmp
Execute hadoop fs commands on the command line in these directories:
•
Packaged or AMI distributions:
•
$ dse hadoop fs option
Tarball installs:
$ install_location/bin/dse hadoop fs option
For example, using this syntax, you can load MapReduce input from the local file system into the
Cassandra File System on Linux.
$ dse hadoop fs -mkdir /user/hadoop/wordcount/input
$ dse hadoop fs -copyFromLocal $HADOOP_EXAMPLE/data/state_of_union/
state_of_union.txt
/user/hadoop/wordcount/input
To list all options for performing command hadoop HDFS commands:
$ dse hadoop fs -help
The DSE command reference lists other commands.
Managing the job tracker using dsetool commands
Examples for using dsetool commands to identify and manage job tracker nodes.
About this task
Several dsetool command are useful for managing job tracker nodes:
•
dsetool jobtracker
55
DSE Analytics with Hadoop
•
Returns the job tracker hostname and port to your location in the data center where you issued the
command.
dsetool movejt data center-workload node IP
•
Moves the job tracker and notifies the task tracker nodes.
dsetool movejt node IP
•
If you do not specify the data center name, the command moves the reserve job tracker.
dsetool listjt
•
Lists all job tracker nodes grouped by their local data center.
dsetool ring
Lists the nodes and types of the nodes in the ring.
More dsetool commands and options are described later.
Listing job trackers example
If you are not sure which nodes in your DSE cluster are job trackers, run the following command:
•
Packaged installs:
•
$ dsetool jobtracker
Tarball installs:
$ install_location/bin/dsetool jobtracker
Moving the job tracker node example
If your primary job tracker node fails, you can use dsetool movejt to move the job tracker to another
analytics node in the cluster. In progress MapReduce jobs fail when you move the job tracker node or
when the node goes down.
Procedure
1. Log in to a DataStax Enterprise analytics node.
2. Run the dsetool movejt command and specify the data center name, hyphen, Analytics (for the
workload), and the IP address of the new job tracker node in your DataStax Enterprise cluster. For
example, to move the job tracker to node 110.82.155.4 in the DC1 data center:
•
Packaged installation:
•
$ dsetool movejt DC1-Analytics 110.82.155.4
Tarball installs:
$ install_location/bin/dsetool movejt DC1-Analytics 110.82.155.4
3. Allow 20 seconds for all of the analytics nodes to detect the change and restart their task tracker
processes.
4. In a browser, connect to the new job tracker and confirm that it is up and running. For example (change
the IP to reflect your job tracker node IP):
http://110.82.155.4:50030
5. If you are running Hive or Pig MapReduce clients, you must restart them to pick up the new job tracker
node information.
Changing the job tracker client port
The job tracker listens on a port for client messages. To use a port other than the default port 8012,
configure the mapred.job.tracker property.
56
DSE Analytics with Hadoop
About this task
By default, the job tracker listens on port 8012 for client messages. You can use another port by
configuring the mapred.job.tracker property.
Procedure
1. Open the mapred-site.xml file for editing. The location of the file depends on the type of installation:
• Packaged installation: /etc/dse/hadoop
• Tarball installs: install_location/resources/hadoop/conf
2. Locate the mapred.job.tracker property:
<!-- Auto detect the dse job tracker -->
<property>
<name>mapred.job.tracker</name>
<value>${dse.job.tracker}</value>
<description>
The address of the job tracker
</description></pre></stepxmp>
3. In the mapred.job.tracker property, change the placeholder ${dse.job.tracker} value to the port number
you want to use. For example, change the port number from the default to 8013:
<!-- Auto detect the dse job tracker -->
<property>
<name>mapred.job.tracker</name>
<value>8013</value>
<description>
The address of the job tracker
</description>
About the Cassandra File System
A Hive or Pig analytics job requires a Hadoop file system to function. For use with DSE Hadoop, DataStax
Enterprise provides a replacement for the Hadoop Distributed File System (HDFS) called the Cassandra
File System (CFS).
About this task
A Hive or Pig analytics job requires a Hadoop file system to function. DataStax Enterprise provides
a replacement for the Hadoop Distributed File System (HDFS) called the Cassandra File System
(CassandraFS), which serves this purpose. When an analytics node starts up, DataStax Enterprise creates
a default CassandraFS rooted at cfs:/ and an archive file system named cfs-archive.
Configuring a CFS superuser
A CFS superuser is the DSE daemon user, the user who starts DataStax Enterprise. A cassandra
superuser, set up using the CQL CREATE USER command, is also a CFS superuser.
A CFS superuser can modify files in the CassandraFS without any restrictions. Files that a superuser adds
to the CassandraFS are password-protected.
Deleting files from the Cassandra File System
Cassandra does not immediately remove deleted data from disk when you use the dse hadoop fs rm file command. Instead, Cassandra treats the deleted data like any data deleted from Cassandra. A
tombstone is written to indicate the new data status. Data marked with a tombstone exist for a configured
time period (defined by the gc_grace_seconds value set on the table). When the grace period expires, the
compaction process permanently deletes the data. You do not have to manually remove expired data.
57
DSE Analytics with Hadoop
Using multiple Cassandra File Systems
DataStax Enterprise supports multiple CassandraFS's. Some typical reasons for using an additional
CassandraFS are:
•
•
•
•
To isolate hadoop-related jobs
To configure keyspace replication by job
To segregate file systems in different physical data centers
To separate Hadoop data in some other way
Creating multiple Cassandra File Systems
Procedure
1. Open the core-site.xml file for editing. The location of the file depends on your installation:
• Tarball installs: /etc/dse/hadoop
• Packaged installation: install_location/resources/hadoop/conf
2. Add one or more property elements to core-site.xml using this format:
<property>
<name>fs.cfs-<filesystemname.impl></name>
<value>com.datastax.bdp.hadoop.cfs.CassandraFileSystem</value>
</property>
3. Save the file and restart Cassandra.
DSE creates the new CassandraFS.
4. To access the new CassandraFS, construct a URL using the following format:
cfs-<filesystemname>:<path>
For example, assuming the new file system name is NewCassandraFS use the dse commands to put
data on the new CassandraFS.
dse hadoop fs -put /tmp/giant_log.gz cfs-NewCassandraFS://cassandrahost/tmp
dse hadoop fs distcp hdfs:/// cfs-NewCassandraFS:///
Using the cfs-archive to store huge files
The Cassandra File System (CFS) consists of two layers: cfs and cfs-archive. Using cfs-archive is
recommended for long-term storage of huge files.
About this task
The CassandraFS consists of two layers, cfs and cfs-archive that you access using these Hadoop shell
commands and URIs:
•
•
cfs:// for the cassandra layer
cfs-archive:// for the cassandra archive layer
Using cfs-archive is highly recommended for long-term storage of huge files, such as those having
terabytes of data. On the contrary, using cfs is not recommended because the data on this layer undergoes
the compaction process periodically, as it should. Hadoop uses the cfs layer for many small files and
temporary data, which need to be cleaned up after deletions occur. When you use the cfs layer instead
of the cfs-archive layer, compaction of huge files can take too long, for example, days. Files stored on the
cfs-archive layer, on the other hand, do not undergo compaction automatically. You can manually start
compaction using the nodetool compact command.
58
DSE Analytics with Hadoop
Example: Store a file on cfs-archive
This example shows how to store a file on cfs-archive using the Hadoop shell commands from the
DataStax Enterprise installation directory on Linux:
1. Create a directory on the cfs-archive layer. You need to use an additional forward slash, as described
earlier:
bin/dse hadoop fs -mkdir cfs-archive:///20140401
2. Use the Hadoop shell put command and an absolute path name to store the file on the cfs-archive
layer.
bin/dse hadoop fs -put big_archive.csv cfs-archive:///20140401/
big_archive.csv
3. Check that the file is stored in on the cfs-archive.
bin/dse hadoop fs -ls cfs-archive:///20140401/
Example: Migrate a file from SQL to text on cfs-archive
This example shows how to migrate the data from the MySQL table the archive directory cfs-archive/
npa_nxx.
1. Run the sqoop demo.
2. Use the dse command in the bin directory to migrate the data from the MySQL table to text files in the
npa_nxx directory of cfs-archive. Specify the IP address of the host in the --target-dir option.
$ sudo ./dse sqoop import --connect
jdbc:mysql://127.0.0.1/npa_nxx_demo
--username root
--password <password>
--table npa_nxx
--target-dir cfs-archive://127.0.0.1/npa_nxx
Using Hive
DataStax Enterprise includes a Cassandra-enabled Hive MapReduce client.
About this task
DataStax Enterprise includes a Cassandra-enabled Hive MapReduce client. Hive is a data warehouse
system for Hadoop that projects a relational structure onto data stored in Hadoop-compatible file systems.
You query the data using a SQL-like language called HiveQL.
You can start the Hive client on any analytics node and run MapReduce queries directly on data already
stored in Cassandra. Using the DataStax Enterprise ODBC driver for Hive, a JDBC compliant user
interface can connect to Hive from the Hive server.
Why Hive
By using Hive, you typically eliminate boilerplate MapReduce code and enjoy productivity gains. The
large base of SQL users can master HiveQL quickly. Hive has a large set of standard functions, such as
mathematical and string functions. You can use Hive for queries that Cassandra as a NoSQL database
does not support, such as joins. DataStax Enterprise support of Hive facilitates the migration of data to
DataStax Enterprise from a Hive warehouse. Hive capabilities are extensible through a Hive user-defined
function (UDF), which DataStax Enterprise supports.
Typical uses for Hive are:
•
Reporting
User engagement and impression click count applications
59
DSE Analytics with Hadoop
•
•
Ad hoc analysis
Machine learning
Advertising optimization
Hive in DataStax Enterprise
DataStax Enterprise analytics nodes store Hive table structures in the CassandraFS instead of in a
Hadoop Distributed File System (HDFS). You layer a Hive table definition onto a directory in the file system
or onto a Cassandra CQL table. The Hive table definition describes the layout of the data and is stored in
the HiveMetaStore keyspace. DataStax Enterprise implements the Hive metastore as the HiveMetaStore
keyspace within Cassandra. Unlike open source Hive, there is no need to run the metastore as a
standalone database to support multiple users.
The consistency level of Hadoop nodes is ONE by default, but when processing Hive queries, if DataStax
Enterprise can guarantee that a replicas are in the same data center, the consistency level of LOCAL_ONE
is used.
You can map an existing CQL table to Hive a table by creating an external table. The external table data
source is external to Hive, located in CQL. When you drop a Hive external table, only the table metadata
stored in the HiveMetaStore keyspace is removed. The data persists in CQL.
Instead of an external table, you can use a Hive managed table. Hive manages storing and deleting the
data in this type of table. The data source can be a flat file that you put on the CassandraFS (using a
hadoop -fs command) or the file can be elsewhere, such as on an operating system file system. To load
the file, use the LOAD [LOCAL] DATA INPATH, INSERT INTO, or INSERT OVERWRITE Hive commands.
Hive metastore configuration
The HiveMetaStore in DataStax Enterprise supports multiple users and requires no configuration except
increasing the default replication factor of the keyspace. The default replication for system keyspaces is 1.
This replication factor is suitable for development and testing, not for a production environment. To avoid
production problems, alter the replication factor of these system keyspaces from 1 to a higher number.
•
•
•
HiveMetaStore
cfs
cfs_archive keyspaces
To prevent missing data problems or data unavailable exceptions after altering keyspaces that contain any
data, run nodetool repair as shown in these examples.
60
DSE Analytics with Hadoop
Supported Hive features
The Hive component in DataStax Enterprise 4.0 has been upgraded to Hive 0.12. The new version of Hive
includes new querying capabilities, data type mapping, and performance enhancements. The following
Hive 0.12 features are supported:
•
Windowing functions
•
• RANK
• LEAD/LAG
• ROW_NUMBER
• FIRST_VALUE, LAST_VALUE
Aggregate OVER functions with PARTITION BY and ORDER BY
DataStax Enterprise supports most CQL and Cassandra internal data types. DataStax provides a Hive
user-defined function (UDF) for working with unsupported types, such as blob:
org.apache.hadoop.hive.cassandra.ql.udf.UDFStringToCassandraBinary
This UDF converts from Hive Strings to native Cassandra types. Due to limitations in Hive, the UDF can be
used only to convert Hive Strings to string primitives, not collections that are arrays and maps of strings. It
is not possible to use the UDF to convert, for example, an array of strings representing inet addresses to
an array of InetAddress columns in Cassandra.
CQL collections and the CQL composite partition keys are now supported.
The Hive examples in this documentation show how to use new features:
•
•
•
How to use the UDF to view blob output
How to use a CQL collection set
How to use a composite partition key
There are also examples of using external and managed tables.
Running Hive
Run Hive as a server or as a client. HiveServer is an optional service for remote clients to submit
programmatic requests to Hive.
About this task
You can run Hive as a server or as a client. Use a Hive client on a node in the cluster under these
conditions:
•
•
To connect to the Hive server running on another node
To use Hive in a single-node cluster
Start a Hive client
You can start a Hive client on any analytics node and run MapReduce queries directly on data already
stored in Cassandra. You run Hive as a client to perform the examples in this document.
Procedure
1. Start DataStax Enterprise as an analytics (Hadoop) node.
•
Packaged installs:
1. Enable Hadoop mode by setting this option in /etc/default/dse:
HADOOP_ENABLED=1
2. Use this command to start the service:
•
$ sudo service dse start
Tarball installs:
61
DSE Analytics with Hadoop
From the installation directory:
$ bin/dse cassandra -t
2. Start a Hive client.
•
Packaged installs:
•
$ dse hive
Tarball installs:
$ install_location/bin/dse hive
The hive prompt appears and you can now enter HiveQL shell commands.
Browsing through Cassandra tables in Hive
If a keyspace and table exists in Cassandra, you can query the keyspace and table in Hive.
If a keyspace exists in Cassandra, you can use the keyspace in Hive. For example, create a keyspace
in Cassandra using cqlsh. Add some data to the table using cqlsh, and then access the data in Hive by
simply issuing a USE command on the Hive command line.
cqlsh> CREATE KEYSPACE cassandra_keyspace WITH replication =
{'class': 'NetworkTopologyStrategy', 'Analytics': 1};
cqlsh> USE cassandra_keyspace;
cqlsh:cassandra_keyspace> CREATE TABLE exampletable
( key int PRIMARY KEY , data text );
cqlsh:cassandra_keyspace> INSERT INTO exampletable (key, data )
VALUES ( 1, 'This data can be read
automatically in hive');
cqlsh:cassandra_keyspace> quit;
At this point, you can start Hive and use the keyspace in Hive.
hive> USE cassandra_keyspace;
hive> SHOW TABLES;
OK
exampletable
hive> SELECT * FROM exampletable;
OK
1 This data can be read automatically in hive
Creating or altering CQL data from Hive
Use a Hive external table to create or alter CQL data from Hive. A counterpart to the Hive database/
external table must pre-exist in Cassandra as an keyspace/table.
You need to use a Hive external table to create or alter CQL data from Hive. A counterpart to the Hive
database/external table must pre-exist in Cassandra as an keyspace/table. When you use a Hive database
name that matches a Cassandra keyspace name, DataStax Enterprise 4.0.4 and later automatically
generates a Hive external table for each table in the keyspace. If the auto-created external table does not
suit your needs, you create a custom external table using different TBL and SERDEPROPERTIES. Use
the CREATE EXTERNAL TABLE statement to create such a table.
To use Hive with legacy tables, such as those created using Thrift or the CLI, see DataStax Enterprise 3.0
documentation. Thrift applications require that you configure Cassandra for connection to your application
using the rpc connections instead of the default native_transport connection.
Creating a custom external table
This example assumes you created the cassandra_keyspace and exampletable in "Browsing Cassandra
tables in Hive". A Hive example table is auto-created when you run the USE cassandra_keyspace
command on the Hive command line. If you want to use a Hive database or table of a different name
than the auto-created ones, but with the same or a similar schema, you customize the auto-created
62
DSE Analytics with Hadoop
external table, as shown in this example. The example uses the Hive database named bigdata instead
cassandra_keyspace, and the example uses a table named MyHiveTable instead of exampletable.
The example specifies the CQL keyspace and table names in the external table definition using the
TBLPROPERTIES clause to use the CQL-defined schema.
Creating a custom external table
hive> CREATE DATABASE bigdata;
hive> USE bigdata;
hive> CREATE EXTERNAL TABLE MyHiveTable
( key int, data string )
STORED BY 'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
TBLPROPERTIES ( "cassandra.ks.name" = "cassandra_keyspace" ,
"cassandra.cf.name" = "exampletable" ,
"cassandra.ks.repfactor" = "2" ,
"cassandra.ks.strategy" =
"org.apache.cassandra.locator.SimpleStrategy" );
Updating metadata in Hive when altering tables
When you run ALTER TABLE, the metadata in Hive is not updated and subsequent Hive and SparkSQL
queries fail.
Workaround
1. Enter the hive shell:
$ dse hive
2. In the hive shell, drop the table:
hive> DROP TABLE your_keyspace.your_table;
3. To allow Hive to refresh the metadata:
hive> USE your_keyspace;
Hive to Cassandra type mapping
In the previous version of DataStax Enterprise, the Hive date, decimal and varchar data types were not
supported. The following table, which shows CQL, Cassandra internal storage engine (used by legacy
tables), and Hive data type mapping, includes these Hive types and mapping information:
CQL
Cassandra Internal
Hive
ascii
AsciiType
string
bigint
LongType
bigint
boolean
BooleanType
boolean
counter
CounterColumnType
bigint
decimal
DecimalType
decimal (new)
double
DoubleType
double
float
FloatType
float
inet
InetAddressType
binary
int
Int32Type
int
text
UTF8Type
string
63
DSE Analytics with Hadoop
CQL
Cassandra Internal
Hive
timestamp
TimestampType
date (new)
timestamp
TimestampType
timestamp (change)
timeuuid
TimeUUIDType
binary
uuid
UUIDType
binary
varint
IntegerType
binary
varchar
UTF8Type
varchar (new)
other
other
binary
The InetAddressType stores the raw IP address in network byte order.
Using TBLPROPERTIES and SERDEPROPERTIES
In an external table definition, the TBLPROPERTIES clause maps the Hive database to a CQL table and
can include MapReduce properties, Cassandra database configuration, and native protocol properties for
the table. The SERDEPROPERTIES clause specifies the properties used when serializing/deserializing
data passed between the Hive table and Cassandra. You can add a WITH SERDEPROPERTIES clause to
map meaningful column names in Hive to the Cassandra partition key, column names, and column values.
You can change these properties on the fly. Using the Hive SET command, you can configure properties in
the hive session. The settings become effective for the next query.
The following table lists general properties used in the TBLPROPERTIES or SERDEPROPERTIES clause
or both. The next section lists optional native protocol properties for use with the DataStax Java Driver.
Table 1: TBL and SERDE properties
64
General Property
TBL/SERDE
Description
cassandra.cf.name
both
Cassandra table name
cassandra.columns.mapping both
Mapping of Hive to legacy Cassandra columns
cassandra.consistency.level both
Consistency level - default ONE
cassandra.cql3.type
both
CQL types
cassandra.host
both
IP of a Cassandra node to connect to
cassandra.input.split.size
both
MapReduce split size
cassandra.ks.name
both
Cassandra keyspace name
cassandra.page.size
SERDE
Fetch tables with many columns by page size
cassandra.partitioner
both
Partitioner (default = configured partitioner)
cassandra.port
both
Cassandra RPC port - default 9160
cql3.output.query
TBL
A prepared statement for storing alterations to a CQL
users table
cql3.partition.key
both
CQL partition key, a comma-separated list of partition
and clustering keys (DataStax 4.0.4 and later)
cql3.pushdown.enable
TBL
True (default) enable pushdown predicate
cql3.update.columns
both
Used with INSERT INTO SELECT
DSE Analytics with Hadoop
Required table properties
When you create an external table in Hive, you need to specify these properties:
•
•
cassandra.ks.name
cassandra.cf.name
Other frequently-used properties are:
•
•
cql3.output.query
cql3.partition.key (DataStax Enterprise 4.0.4 and later)
In DataStax Enterprise 4.0.4 and later, you use the SHOW CREATE TABLE <CQL table name>
command at the Hive prompt to see the auto-created external table. The output helps you see how to
format the cql3.partition.key in your custom external table. For example, the output of a table having
following CQL composite partition key, has the 'cql3.partition.key'='key,event_id' Hive property syntax.
PRIMARY KEY ((key, event_id), num_responses)
Required storage handler
Also required in the external table definition is the CQL storage handler:
org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler. The storage handler accesses and stores
Cassandra data back to Cassandra.
About the cassandra.input.split.size
The cassandra.input.split.size property configures the number of rows processed per mapper (64k rows
per split). The default is 64 * 1024
Partitioner use by Hive
You do not need to specify cassandra.partitioner. Your configured partitioner is used by Hive. For example,
Hive uses this property value if you use the Cassandra 2.0 default partitioner:
"cassandra.partitioner" = "org.apache.cassandra.dht.Murmur3Partitioner"
Creating or altering CQL data from Hive and MapReduce performance show examples of using some of
these properties.
Accessing tables with many columns
In the TBLPROPERTIES clause, set the cassandra.page.size to fetch a table with many columns at once
by page size.
Optional native protocol properties
DataStax Enterprise 4.0.4 and later support the following optional properties for the native protocol.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
cassandra.input.native.port
cassandra.input.native.core.connections.per.host
cassandra.input.native.max.connections.per.host
cassandra.input.native.min.simult.reqs.per.connection
cassandra.input.native.max.simult.reqs.per.connection
cassandra.input.native.connection.timeout
cassandra.input.native.read.connection.timeout
cassandra.input.native.receive.buffer.size
cassandra.input.native.send.buffer.size
cassandra.input.native.solinger
cassandra.input.native.tcp.nodelay
cassandra.input.native.reuse.address
cassandra.input.native.keep.alive
cassandra.input.native.auth.provider
65
DSE Analytics with Hadoop
•
•
•
•
•
cassandra.input.native.ssl.trust.store.path
cassandra.input.native.ssl.key.store.path
cassandra.input.native.ssl.trust.store.password
cassandra.input.native.ssl.key.store.password
cassandra.input.native.ssl.cipher.suites
Inspecting an auto-created, external table (DataStax Enterprise 4.0.4 and later)
In Hive, you can use the SHOW CREATE TABLE <CQL table name> command to see the schema of
a auto-created external table. The output of this command can help you construct a custom Hive external
table definition. Assuming you created the table in "Browsing through Cassandra tables in Hive", use the
SHOW CREATE TABLE command to see the schema of exampletable.
hive> SHOW CREATE TABLE exampletable;
OK
CREATE EXTERNAL TABLE exampletable(
key int COMMENT 'from deserializer',
data string COMMENT 'from deserializer')
ROW FORMAT SERDE
'org.apache.hadoop.hive.cassandra.cql3.serde.CqlColumnSerDe'
STORED BY
'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
WITH SERDEPROPERTIES (
'serialization.format'='1',
'cassandra.columns.mapping'='key,data')
LOCATION
'cfs://127.0.0.1/user/hive/warehouse/cassandra_keyspace.db/exampletable'
TBLPROPERTIES (
'cassandra.partitioner'='org.apache.cassandra.dht.Murmur3Partitioner',
'cql3.partition.key'='key',
'cassandra.ks.name'='cassandra_keyspace',
'cassandra.cf.name'='exampletable',
'auto_created'='true')
Time taken: 0.028 seconds, Fetched: 18 row(s)
The CREATE EXTERNAL TABLE definition uses the cql3.partition.key TBL property, available only in
DataStax Enterprise 4.0.4 and later.
Updating metadata in Hive when altering tables
When you run ALTER TABLE, the metadata in Hive is not updated and subsequent Hive and SparkSQL
queries fail.
Workaround
1. Enter the hive shell:
$ dse hive
2. In the hive shell, drop the table:
hive> DROP TABLE your_keyspace.your_table;
3. To allow Hive to refresh the metadata:
hive> USE your_keyspace;
Using a managed table to load local data
If you do not need to store data in a Cassandra table, use a managed table instead of an external table.
The data can be located in the Cassandra File System (CFS) or on the file system.
66
DSE Analytics with Hadoop
If you do not need to store data, use a managed table instead of an external table. The data can be located
in the CassandraFS or on the file system. You load the data into the managed table as shown in this
example:
1. Create a managed table:
hive>
CREATE TABLE invites (foo INT, bar STRING )
PARTITIONED BY (ds STRING );
2. Load data into a table using the LOAD DATA command. The HiveQL Manual provides more information
about the HiveQL syntax.
For example, on the Mac OS X:
hive>
LOAD DATA LOCAL
INPATH '<install_location>/resources/hive/examples/files/kv2.txt'
OVERWRITE INTO TABLE invites PARTITION ( ds = '2008-08-15' );
hive>
LOAD DATA LOCAL
INPATH '<install_location>/resources/hive/examples/files/kv3.txt'
OVERWRITE INTO TABLE invites PARTITION ( ds = '2008-08-08' );
hive>
SELECT count (* ), ds FROM invites GROUP BY ds;
Note: The paths to the Hive example files shown in the example LOAD commands above are for
the tarball distribution.
Using an external file system
You can map a file in an external file system, such as S3 native file system to a table in Hive.
About this task
You can map a file in an external file system, such as S3 native file system to a table in Hive. The DSE
Hadoop cluster continues to use the Cassandra File System (CFS) file system. The data source is external
to Hive, located in S3 for example. You create a Hive external table for querying the data in an external file
system. When you drop the external table, only the table metadata stored in the HiveMetaStore keyspace
is removed. The data persists in the external file system.
First, you set up the hive-site.xml and core-site.xml files, and then create an external table as
described in this procedure.
Procedure
1. Open the hive-site.xml for editing. This file is located in:
•
Packaged installations: /etc/dse/hadoop
67
DSE Analytics with Hadoop
• Tarball installations: install_location/resources/hive/conf
2. Add a property to hive-site.xml to set the default file system to be the native S3 block file system.
Use fs.default.name as the name of the file system and the location of the bucket as the value. For
example, if the S3 bucket name is mybucket:
<property>
<name>fs.default.name</name>
<value>s3n://mybucket</value>
</property>
3. Save the file.
4. Open the core-site.xml file for editing. This file is located in:
• Packaged installations: /etc/dse/hadoop
• Tarball installations: install_location/resources/hadoop/conf
5. Add these properties to core-site.xml to specify the access key ID and the secret access key credentials
for accessing the native S3 block filesystem:
<property>
<name>fs.s3n.awsAccessKeyId</name>
<value>ID</value>
</property>
<property>
<name>fs.s3n.awsSecretAccessKey</name>
<value>Secret</value>
</property>
6. Save the file and restart Cassandra.
7. Create a directory in s3n://mybucket named, for example, mydata_dir.
8. Create a data file named mydata.txt, for example. Delimit fields using =.
"key1"=100
"key2"=200
"key3"=300
9. Put the data file you created in s3n://mybucket/mydata_dir.
10.Using cqlsh, create a keyspace and a CQL table schema to accommodate the data on S3.
cqlsh> CREATE KEYSPACE s3_counterpart WITH replication =
{'class': 'NetworkTopologyStrategy', 'Analytics': 1};
cqlsh> USE s3_counterpart;
cqlsh:s3_counterpart> CREATE TABLE mytable
( key text PRIMARY KEY , data int );
11.Start Hive, and on the Hive command line, create an external table for the data on S3. Specify the S3
file name as shown in this example.
hive> CREATE EXTERNAL TABLE mytable (key STRING, value INT) ROW FORMAT
DELIMITED FIELDS TERMINATED BY '=' STORED AS TEXTFILE LOCATION 's3n://
mybucket/mydata_dir/';
Now, having the S3 data in Hive, you can query the data using Hive.
12.Select all the data in the file on S3.
SELECT * from mytable;
OK
key1 100
key2 200
key3 300
Example: Work with an unsupported data type
DataStax Enterprise provides a user defined function (UDF) for converting Hive binary data into string
representations of CQL types.
68
DSE Analytics with Hadoop
DataStax Enterprise provides a user defined function (UDF) for converting Hive binary data into string
representations of CQL types. Basically the data in Hive is seen as binary because it doesn't support
all the various types we support in Cassandra. Because these types are stored as binary, if you want to
actually read the data in Hive in a human comprehensible format, the data must be converted using the
provided UDF.
Create the keyspace and two tables in cqlsh
This example first creates a keyspace and two tables in cqlsh and inserts data of every supported type into
the tables.
1. Start cqlsh. For example, on Linux.
./cqlsh
2. In cqlsh, create a keyspace:
cqlsh> CREATE KEYSPACE cql3ks WITH replication =
{ 'class': 'NetworkTopologyStrategy',
'Analytics': '1' };
3. Using cqlsh, create a table in the cql3ks keyspace having columns of every CQL data type.
cql> USE cql3ks;
cql3ks> CREATE TABLE genericAdd (
key ascii PRIMARY KEY, a bigint, b blob, c boolean,
d decimal, e double, f float, g inet, h int, i text,
j timestamp, k uuid, l timeuuid, m varint);
4. Insert some data into the table.
cql3ks> INSERT INTO genericAdd (
key,a,b,c,d,e,f,g,h,i,j,k,l,m)
VALUES ('KeyOne', 100005, 0xBEEFFEED, true, 3.5,-1231.4,
3.14, '128.2.4.1', 42, 'SomeText', '2008-10-03',
e3d81c40-1961-11e3-8ffd-0800200c9a66,
f078d660-1961-11e3-8ffd-0800200c9a66, 1000000);
5. Create a second table, genericToAdd, containing every data type and insert different data into the table.
cql3ks> CREATE TABLE genericToAdd (
id int PRIMARY KEY, key ascii, a bigint, b blob, c boolean,
d decimal, e double, f float, g inet, h int, i text,
j timestamp, k uuid, l timeuuid, m varint);
6. Insert some data into the second table.
cql3ks> INSERT INTO genericToAdd (
id,key,a,b,c,d,e,f,g,h,i,j,k,l,m)
VALUES (1,'Oneness',1, 0x11111111, true, 1.11,-1111.1,1.11,
'111.1.1.1', 11,'11111','1999-11-01',
e3d81c40-1961-11e3-8ffd-0800200c9a66,
f078d660-1961-11e3-8ffd-0800200c9a66, 1);
Create an external table in Hive
Next, create an external table in Hive that maps to a pre-existing CQL 3 table in Cassandra. You cannot
use the auto-created table (DataStax Enterprise 4.0.4 and later only) because Hive cannot represent the
blob type in a comprehensible format. After creating the custom external table, you can perform alterations
of the CQL tables from Hive. You insert data from the second CQL table into the first CQL table from Hive.
Using a UDF, you query the external table in Hive. You need to use the UDF because the data is of the
unsupported blob type.
DataStax Enterprise 4.0.4 and later
1. Create a table in Hive that includes a cql3.output.query property that has the value of a prepared
statement for inserting the data from the second, genericToAdd, table into the first, genericAdd, table.
69
DSE Analytics with Hadoop
The last couple of lines in the following statement need to be free of line breaks. If you copy/paste
this statement directly from the documentation and do not remove line breaks, an error occurs in the
subsequent step.
DataStax Enterprise 4.0.4 and later
hive> CREATE EXTERNAL TABLE hive_genericadd (
key string, a bigint, b binary, c boolean, d decimal,
e double, f float, g binary, h int, i string, j timestamp,
k binary, l binary, m binary) STORED BY
'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
TBLPROPERTIES ( "cassandra.ks.name" = "cql3ks",
"cassandra.cf.name" = "genericadd",
"cql3.partition.key"="key",
"cql3.output.query" =
"INSERT%20INTO%20cql3ks.genericadd%20
(key%2Ca%2Cb%2Cc%2Cd%2Ce%2Cf%2Cg%2Ch%2Ci%2Cj%2Ck%2Cl%2Cm)
%20VALUES%20
(%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F
%2C%3F)");
DataStax Enterprise 4.0 - 4.0.3
hive> CREATE EXTERNAL TABLE hive_genericadd (
key string, a bigint, b binary, c boolean, d decimal,
e double, f float, g binary, h int, i string, j timestamp,
k binary, l binary, m binary) STORED BY
'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
TBLPROPERTIES ( "cassandra.ks.name" = "cql3ks",
"cassandra.cf.name" = "genericadd",
"cql3.output.query" =
"INSERT%20INTO%20cql3ks.genericadd%20
(key%2Ca%2Cb%2Cc%2Cd%2Ce%2Cf%2Cg%2Ch%2Ci%2Cj%2Ck%2Cl%2Cm)
%20VALUES%20
(%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F%2C%3F
%2C%3F)");
2. Use the INSERT statement to start the MapReduce job that inserts the data from the second CQL table
into the first one.
hive> INSERT INTO TABLE hive_genericadd SELECT key,a,b,c,d,e,f,g,h,i,j,k,l,m
FROM cql3ks.generictoadd;
The MapReduce job runs.
Total MapReduce jobs = 1
Launching Job 1 out of 1
. . .
Job 0: Map: 2
HDFS Read: 0 HDFS Write: 0 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
Time taken: 33.278 seconds
3. Create an alias for the UDF provided by DataStax.
hive> CREATE TEMPORARY FUNCTION c_to_string AS
'org.apache.hadoop.hive.cassandra.ql.udf.UDFCassandraBinaryToString';
4. Select the data of the unsupported blob type from the Hive table by calling the UDF.
hive> select c_to_string(b, 'blob') from hive_genericadd;
The MapReduce job runs, and the output correctly displays the values:
Total MapReduce jobs = 1
. . .
70
DSE Analytics with Hadoop
Job 0: Map: 2
HDFS Read: 0 HDFS Write: 0 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
beeffeed
11111111
INSERT INTO SELECT statement
DataStax Enterprise supports the INSERT INTO SELECT statement in Hive.
About this task
DataStax Enterprise supports the INSERT INTO SELECT statement in Hive. You set a TBL and SERDE
property, and use INSERT INTO SELECT to copy data from one table and insert it into another, or the
same, table.
Supported TBL and SERDE properties include the following new SERDE property:
cql3.update.columns
You use cql3.update.columns in conjunction with the CQL output query property, cql3.output.query.
The following example shows how to configure these properties and use the INSERT INTO SELECT
statement in Hive to insert selective columns from a table into another row of the same Cassandra table.
The SELECT statement requires values for each column in the target table. Using fake values satisfies this
requirement.
Procedure
1. Start cqlsh and create a Cassandra keyspace and table.
cqlsh> CREATE KEYSPACE mykeyspace WITH replication = {'class':
'SimpleStrategy', 'replication_factor': 3};
cqlsh> USE mykeyspace;
cqlsh> CREATE TABLE mytable (a INT PRIMARY KEY, b INT, c INT, d INT);
cqlsh> INSERT INTO mytable (a, b, c, d) VALUES (1, 2, 3, 4);
2. Start the Hive client.
3. In Hive, create a Hive external table for altering the Cassandra table you just created.
hive> CREATE EXTERNAL TABLE mytable (a int, b int, c int, d int)
STORED BY 'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
TBLPROPERTIES ( "cassandra.ks.name" = "mykeyspace",
"cassandra.cf.name" = "mytable",
"cassandra.ks.repfactor" = "3",
"cassandra.ks.strategy" =
"org.apache.cassandra.locator.SimpleStrategy");
4. In Hive, select all the data in mytable.
hive> SELECT * from mytable;
Output is:
OK 1 2 3 4 Time taken: 0.138 seconds, Fetched: 1 row(s)
5. In Hive, alter the external table to configure the prepared statement as the value of the Hive CQL output
query. The prepared statement in this example takes values inserted into columns a and b in mytable
and maps them to columns b and a, respectively, for insertion into the new row.
hive> ALTER TABLE mytable SET TBLPROPERTIES ('cql3.output.query' = 'update
mykeyspace.mytable set b = ? where a = ?');
hive> ALTER TABLE mytable SET SERDEPROPERTIES ('cql3.update.columns' =
'b,a');
6. In Hive, execute an INSERT INTO SELECT statement to insert a row of data into mytable. For
example, use 4 and 9 as the values to insert into the first two positions (a, b) of the row. The CQL
71
DSE Analytics with Hadoop
output query will reverse these positions. Use two type-compatible fake values in addition to the values
4 and 9 that you want to insert. In this example, the fake values are an int, 9999, and a column name, d.
hive> INSERT INTO TABLE mytable SELECT 4, 9, 9999, d FROM mytable;
The MapReduce job runs:
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks is set to 0 since there's no reduce operator
. . .
MapReduce Jobs Launched:
Job 0: Map: 2
HDFS Read: 0 HDFS Write: 0 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
Time taken: 31.867 seconds
7. Check that 4 and 9, and only those values, were inserted:
hive> SELECT * FROM mytable;
The fake values are inserted as NULL and only the values specified by the
CQL output query are inserted. The output query mapped 4 to column b and 9
to column a.
OK
1
2
3
4
9
4
NULL
NULL
Time taken: 0.131 seconds, Fetched: 2 row(s)
Example: Use a CQL composite partition key
Example steps to create a CQL table, and then create an external table in Hive that maps to the CQL table.
About this task
This example first creates a CQL table, and then creates an external table in Hive that maps to the CQL
table. The Hive table uses a SERDE property and declares a single key followed by the number of column
declarations that match the number of columns in the CQL table. Finally, the example queries the CQL
table from Hive.
Procedure
1. In cqlsh, add a table to the cql3ks keyspace created earlier. Create a table that uses a composite
partition key.
cql3ks> CREATE TABLE event_table (
key ascii, factor float, event_type text, event_date timestamp,
event_id timeuuid, num_responses varint,
PRIMARY KEY ((key, event_id), num_responses)
);
2. Insert data into the table.
cql3ks> INSERT INTO event_table (
key, factor, event_type, event_date, event_id, num_responses)
VALUES ('KeyOne', 3.14, 'Q3-launch', '2014-09-03',
f078d660-1961-11e3-8ffd-0800200c9a66, 1000000
);
3. Create the external table in Hive.
•
DataStax Enterprise 4.0.4 and later
hive> CREATE EXTERNAL TABLE mapped_table(
key string, factor float, event_type string,
event_date timestamp, event_id binary, num_responses binary)
STORED BY 'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
72
DSE Analytics with Hadoop
WITH SERDEPROPERTIES( "cassandra.ks.name" = "cql3ks",
"cassandra.cf.name" = "event_table",
"cql3.partition.key" = "key,event_id",
"cassandra.cql3.type" = "ascii, float, text, timestamp, timeuuid,
varint"
);
•
DataStax Enterprise 4.0 - 4.0.1
hive> CREATE EXTERNAL TABLE mapped_table(
key string, factor float, event_type string,
event_date timestamp, event_id binary, num_responses binary)
STORED BY 'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
WITH SERDEPROPERTIES( "cassandra.ks.name" = "cql3ks",
"cassandra.cf.name" = "event_table",
"cassandra.cql3.type" = "ascii, float, text, timestamp, timeuuid,
varint"
);
4. Trigger a MapReduce job to query the table in Hive.
hive> SELECT COUNT(*) FROM mapped_table;
The output is:
Total MapReduce jobs = 1
Launching Job 1 out of 1
. . .
MapReduce Jobs Launched:
Job 0: Map: 2 Reduce: 1
HDFS Read: 0 HDFS Write: 0 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
1
Time taken: 39.929 seconds
Using CQL collections
Hive supports writing to CQL table collections.
About this task
Hive supports writing to CQL tables, including tables of collections. To store data to a CQL table from Hive,
use prepared statements as shown in these examples:
Prepared statements for a list
UPDATE users SET top_places = ? where user_id = ?
UPDATE users SET top_places = [ 'rivendell', 'rohan' ] WHERE user_id =
'frodo';
UPDATE users SET top_places = ? + top_places where user_id = ?
UPDATE users SET top_places = [ 'the shire' ] + top_places WHERE user_id =
'frodo';
UPDATE users SET top_places = top_places - ? where user_id = ?
UPDATE users SET top_places = top_places - ['riddermark'] WHERE user_id =
'frodo';
About this task
Prepared statement for a map
Prepared statements for a set are similar to those for a list.
UPDATE users SET todo = ? where user_id = ?
UPDATE users
SET todo = { '2012-9-24' : 'enter mordor',
73
DSE Analytics with Hadoop
'2012-10-2 12:00' : 'throw ring into mount doom' }
WHERE user_id = 'frodo';
The following queries are handled as a regular value instead of tuples:
UPDATE users SET top_places[2] = ? where user_id = ?
UPDATE users SET top_places[2] = 'riddermark' WHERE user_id = 'frodo';
UPDATE users SET todo[?] = ? where user_id = ?
UPDATE users SET todo['2012-10-2 12:10'] = 'die' WHERE user_id = 'frodo';
Example: Alter a set collection
Items in a CQL collection are mapped to the Hive types shown in the Hive to Cassandra type mapping
table. The CQL data types not supported in Hive, such as blob, can be used if you transform the fields of
that type using a DataStax-provided UDF.
In cqlsh, you create two tables that contain a collection sets and insert data into the tables. In Hive, you
create an external table that maps to the first table, and then insert data from the second CQL table to the
first CQL table. Finally, in cqlsh, you query the second CQL table to verify that the insertion was made.
1. In cqlsh, create the users table shown in the CQL documentation that contains a set collection column,
and insert data into the table:
cqlsh> CREATE TABLE cql3ks.users (
user_id text PRIMARY KEY,
first_name text,
last_name text,
emails set <text>
);
cqlsh> INSERT INTO cql3ks.users (user_id, first_name, last_name, emails)
VALUES('frodo', 'Frodo', 'Baggins',
{'[email protected]', '[email protected]'});
2. Create a second table that contains data about actors:
cqlsh> CREATE TABLE cql3ks.actors (
user_id text PRIMARY KEY,
first_name text,
last_name text,
emails set<text>
);
cqlsh> INSERT INTO cql3ks.actors (user_id, first_name, last_name, emails)
VALUES ('ejwood', 'Elijah', 'Wood', {'[email protected]'});
3. In Hive, create a table that maps to the CQL users table. If the table exists in Cassandra, you can
create a corresponding table to access data using the CREATE EXTERNAL TABLE command. The last
couple of lines in the following statement need to be free of line breaks.
hive> CREATE EXTERNAL TABLE hiveUserTable (emails array<string>,user_id
string) STORED BY 'org.apache.hadoop.hive.cassandra.cql3.CqlStorageHandler'
TBLPROPERTIES( "cassandra.ks.name" = "cql3ks", "cassandra.cf.name" =
"users", "cql3.output.query" = "update%20cql3ks.users%20set%20emails%20%3D
%20emails%20%2B%20%3F%20WHERE%20user_id%20%3D%20%3F");
Hive uses a URL-encoded prepared statement to store alterations to the CQL users table. The
output_query specifies the key of the column to alter. The rest of the arguments, the "?"s, for the
prepared statement are filled in by the values related to that key in Hive.
4. Add the data from the CQL actors table to the users table:
hive> INSERT INTO TABLE hiveUserTable SELECT emails,user_id FROM
cql3ks.actors;
The MapReduce job runs and alters the table.
5. Check that the CQL table contains Elijah Wood's email address:
74
DSE Analytics with Hadoop
cql3ks> SELECT * FROM cql3ks.users;
user_id | emails
|
first_name | last_name
---------+-----------------------------------------------------------+------------+----------ejwood |
{[email protected]} |
null |
null
frodo | {[email protected], [email protected], [email protected]} |
Frodo |
Baggins
Creating a Hive CQL output query
One of the Hive external table properties (TBLPROPERTIES) is the cql3.output.query. The value of this
property is a prepared statement that the MapReduce job uses to insert data into the corresponding
Cassandra table.
About this task
One of the Hive external table properties (TBLPROPERTIES) is the cql3.output.query. The value of this
property is a prepared statement that the MapReduce job uses to insert data into the corresponding
Cassandra table. The prepared statement must be url-encoded to make special characters readable by
Hive. The prepared query is identical to the CQL statement for altering the table except the binding of the ?
is done by Hive. The ? are bound to the hive columns in the order specified in the external table schema.
In the example of using a collection set, the external table definition determines the bind variables, '?'s,
needed in the prepared statements:
hive> CREATE EXTERNAL TABLE hiveUserTable
(emails array<string>,user_id string)
. . .
This external table schema specifes the second column to be the user_id; therefore, this INSERT
statement takes the columns emails, user_id from the Cassandra actors table and maps the data into the
Hive emails and user_id columns:
hive> INSERT INTO TABLE hiveUserTable SELECT emails,user_id FROM actors;
The following diagram shows the relationship between the tables and the bind variables:
The hiveUserTable includes this prepared query:
"cql3.output.query" =
"update%20cql3ks.users%20set%20emails%20%3D%20emails
%20%2B%20%3F%20WHERE%20user_id%20%3D%20%3F");
Decoded, the cql3.output.query looks like this:
UPDATE cql3ks.users SET emails = emails + ? WHERE user_id = ?
75
DSE Analytics with Hadoop
The Hive INSERT statement starts the MapReduce job that uses the key value from the actors table in the
'WHERE (user_id) =' clause of prepared statement.
Another example, an abstract one, updates a table having three columns (x,y,z) using a prepared
statement. The query looks like this internally:
create external table ( x,y,z ) Stored by ....
(cql3.output.query = "Update cassX = ?(x) cassY=?(y) where cassZ= ?(z))"
Using a custom UDF
You can include your own Java code in a user-defined function (UDF) and invoke it using a query.
About this task
If the Hive built-in functions do not provide the capability you need, you can include your own Java code
in a user-defined function (UDF) and invoke it using a query. DataStax provides a UDF for working with
unsupported data types, for example. The example in this seciton uses a JAR that converts text from
lowercase to uppercase. After downloading the JAR from the Hadoop tutorial examples repository and
setting up the UDF in Hive, you create a Hive table. You insert data into the table from a text file installed
with DataStax Enterprise. The contents of the file look like this:
238^Aval_238
86^Aval_86
311^Aval_311
27^Aval_27
165^Aval_165
. . .
When you execute a SELECT statement, you invoke the UDF to convert text in the file from lowercase to
uppercase: val to VAL.
Procedure
1. Download the JAR for this example.
2. On the command line, add the JAR to the root Hadoop directory in the CassandraFS using Hadoop
shell commands. For example:
dse hadoop fs -copyFromLocal local-path-to-jar/myudfs.jar /tmp
Substitute the path to the downloaded job in your environment for local-path-to-jar.
3. Start a Hive client, and at the Hive prompt, add the JAR file to the Hadoop distributed cache, which
copies files to task nodes to use when the files run:
hive> add jar cfs:///tmp/myudfs.jar;
The output on the Mac OS X is:
converting to local cfs:///tmp/myudfs.jar
Added /private/tmp/johndoe/hive_resources/myudfs.jar to class path
Added resource: /private/tmp/johndoe/hive_resources/myudfs.jar
4. At the Hive prompt, create an alias for the UDF associated with the JAR.
hive> CREATE TEMPORARY FUNCTION myUpper AS 'org.hue.udf.MyUpper';
5. Create a Hive table for text data.
hive> CREATE TABLE udftest (foo INT, bar STRING);
6. Insert data into the table, substituting the path to the DataStax Enterprise installation in your
environment for the install_location. For example, on Mac OS X:
hive> LOAD DATA LOCAL INPATH
'install_location/resources/hive/examples/files/kv1.txt'
OVERWRITE INTO TABLE udftest;
76
DSE Analytics with Hadoop
7. Convert the lowercase text in the table, the instances of val, to uppercase by invoking the UDF by its
alias in the SELECT statement.
hive> SELECT myUpper(bar) from udftest;
The mapper output looks like this:
. . .
MapReduce Jobs Launched:
Job 0: Map: 1
HDFS Read: 0 HDFS Write: 0 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
VAL_238-gg
VAL_86-gg
VAL_311-gg
. . .
Using pushdown predicates
To minimize the amount of data to be processed, enable pushdown predicates by using
cql3.pushdown.enable in the TBLPROPERTIES clause of a Hive query.
About this task
Pushdown predicates resolve expressions as early as possible in the processing pipeline to minimize
the amount of data to be processed. You enable pushdown predicates using a new property
cql3.pushdown.enable in the TBLPROPERTIES clause of a Hive query. True enables the feature and false
(the default) disables it. Processing of operations on columns of the following types are affected by the
setting:
Cassandra type
Hive type
UTF8Type
string
AsciiType
string
CounterColumnType
long
DateType
timestamp
LongType
long
DoubleType
double
FloatType
float
BooleanType
boolean
Int32Type
int
Recommended usage
When the indexed row is small, enable pushdown predicates; otherwise, disable the feature to avoid a
timeout exception or Out-Of-Memory (OOM) condition.
Limitations
DataStax Enterprise supports pushdown predicates for indexes only. Primary keys are not supported.
Using the Hive count function
For the cassandra.consistency.level general property in Hive TBLPROPERTIES, set the consistency level
to ALL before you issue a Hive SELECT expression that contains the count function.
77
DSE Analytics with Hadoop
About this task
Using the Hive TBLPROPERTIES cassandra.consistency.level, set the consistency level to ALL before
issuing a Hive SELECT expression containing the count function. Using ALL ensures that when you ping
one node for a scan of all keys, the node is fully consistent with the rest of the cluster. Using a consistency
level other than ALL can return resultsets having fewer rows than expected because replication has
not finished propagating the rows to all nodes. A count that is higher than expected can occur because
tombstones have not yet been propagated to all nodes.
To get accurate results from the count function using a consistency level other than ALL:
•
•
Repair all nodes.
Prevent new data from being added or deleted.
Handling schema changes
If you change a table in Cassandra after creating an external table in Hive that maps to that table in
Cassandra, a runtime exception can occur. Use a workaround when changes that occur to the table in
Cassandra get out of synch with the mapped table in Hive.
About this task
If you change a table in Cassandra, using CQL for example, after creating an external table in Hive
that maps to that table in Cassandra, a runtime exception can occur. Changes that occur to the table in
Cassandra get out of synch with the mapped table in Hive. The workaround is:
Procedure
1. In Hive, drop the table.
hive> drop table mytable;
2. Run SHOW TABLES.
hive> show tables;
Now, the table in Hive contains the updated data.
MapReduce performance tuning
DataStax Enterprise includes a Cassandra-enabled Hive MapReduce client. Change settings to enable
improved MadReduce performance.
About this task
You can change performance settings in the following ways:
•
•
•
In an external table definition, using the TBLPROPERTIES or SERDEPROPERTIES clauses.
Using the Hive SET command. For example: SET mapred.reduce.tasks=32;
In the mapred-site.xml file:
•
•
Packaged installs: /etc/dse/hadoop/mapred-site.xml
Tarball installs: install_location/resources/hadoop/conf/mapred-site.xml
Note: This is a system setting so if you change it you must restart the analytics nodes.
Speeding up map reduce jobs
Increase your mappers to one per CPU core by setting mapred.tasktracker.map.tasks.maximum in
mapred-site.xml.
Increasing the number of map tasks to maximize performance
You can increase the number of map tasks in these ways:
78
DSE Analytics with Hadoop
•
•
Turn off map output compression in the mapred-site.xml file to lower memory usage.
The cassandra.input.split.size property specifies rows to be processed per mapper. The default size is
64k rows per split. You can decrease the split size to create more mappers.
Out of Memory Errors
When your mapper or reduce tasks fail, reporting Out of Memory (OOM) errors, turn the
mapred.map.child.java.opts setting in Hive to:
SET mapred.child.java.opts="-server -Xmx512M"
You can also lower memory usage by turning off map output compression in mapred-site.xml.
Using the Fair Scheduler
The Hadoop Fair Scheduler assigns resources to jobs to balance the load, so that each job gets roughly
the same amount of CPU time. The fair-scheduler.xml is located in the resources/hadoop/conf
directory of the DataStax Enterprise installation.
To enable the fair scheduler you uncomment a section in the mapred-site.xml that looks something like
this:
<property>
<name>mapred.jobtracker.taskScheduler</name>
<value>org.apache.hadoop.mapred.FairScheduler</value>
</property>
. . .
<value>dse-3.0.2/dse/resources/hadoop/conf/fair-scheduler.xml</value>
</property>
You might need to change the value element shown here. If the Fair Scheduler file has a different name,
change the name of the file to fair-scheduler.xml. Specify the absolute path to the file.
DataStax Enterprise also supports the Capacity Scheduler.
Starting the Hive server
A node in the analytics cluster acts as the Hive server. To start the Hive server, run the start command
from a node in the Hadoop cluster.
About this task
A node in the analytics cluster can act as the Hive server. Other nodes connect to Hive through the JDBC
driver. To start the Hive server, choose a node in the Hadoop cluster and run this command:
Packaged installs:
$ dse hive --service hiveserver
Tarball installs:
$ install_location/bin/dse hive
--service hiveserver
Starting the HiveServer2
DataStax Enterprise 4.0.4 integrates Apache HiveServer2, an improved version of HiveServer that
supports multi-client concurrency and other features.
To start HiveServer2, run this command:
dse hive --service hiveserver2
After starting HiveServer2, use the Beeline command shell to connect to the server and run Hive queries.
79
DSE Analytics with Hadoop
Using Beeline
DataStax Enterprise supports the HiveServer2 Beeline command shell, a JDBC client. After starting
HiveServer2, open another terminal window, start Beeline, connect to HiveServer2, and run Hive queries.
DataStax Enterprise supports the HiveServer2 Beeline command shell, a JDBC client. HiveServer2, an
improved Hive server, uses Beeline as the command-line interface. After starting HiveServer2, open
another terminal window, start Beeline, connect to HiveServer2, and run Hive queries.
1. In a terminal window, start HiveServer2.
2. In another terminal window, start Beeline. On Linux, for example:
$ install_directory/bin/dse beeline
The beeline prompt appears.
2014-06-19 06:37:22.758 java[46121:1a03] Unable to load realm info from
SCDynamicStore
Beeline version 0.12.0.3-SNAPSHOT by Apache Hive
beeline>
3. Connect to the server. On a single-node, development cluster for example:
beeline> !connect jdbc:hive2://localhost
The HiveServer2 prompt appears.
scan complete in 24ms
Connecting to jdbc:hive2://localhost
Enter username for jdbc:hive2://localhost:
4. Enter the DataStax Enterprise user name.
The password prompt appears.
Enter password for jdbc:hive2://localhost:
5. Enter the password.
The hive2 prompt appears.
Connected to: Hive (version 0.12.0.3-SNAPSHOT)
Driver: Hive (version 0.12.0.3-SNAPSHOT)
Transaction isolation: TRANSACTION_REPEATABLE_READ
0: jdbc:hive2://localhost>
6. Run Hive queries.
Setting the Job Tracker node for Hive
Hive clients automatically select the correct job tracker node upon startup. Use dsetool commands to
configure and manage the job tracker node for an analytics node.
About this task
Hive clients automatically select the correct job tracker node upon startup. You configure and manage the
job tracker node for an analytics node using dsetool commands.
Recreating Hive metadata after decommissioning a node
After removing/decommissioning a node that stored the Hive metadata, truncate the Hive metadata table,
then recreate the table.
After removing/decommissioning a node that stored the Hive metadata, truncate the Hive metadata table,
then recreate the table. In the hive-site.xml, set the parameters as shown in the following example to
specify a different keyspace and table for the Hive metastore:
<property>
<name>cassandra.connection.metaStoreKeyspaceName</name>
<value>newKeyspaceName</value>
</property>
80
DSE Analytics with Hadoop
<property>
<name>cassandra.connection.metaStoreColumnFamilyName</name>
<value>MetaStore</value>
</property>
This action is necessary to prevent an exception in SemanticAnalyzer.genFileSinkPlan.
DataStax ODBC driver for Hive on Windows
The DataStax ODBC Driver for Hive provides Windows users access to the information that is stored in
DSE Hadoop.
About this task
The DataStax ODBC Driver for Hive provides Windows users access to the information stored in the
Hadoop distribution bundled into DataStax Enterprise. This driver allows you to access the data stored
on your DataStax Enterprise Hadoop nodes using business intelligence (BI) tools, such as Tableau
and Microsoft Excel. The driver is compliant with the latest ODBC 3.52 specification and automatically
translates any SQL-92 query into HiveQL.
Before you begin
•
•
•
Windows® 7 Professional or Windows® 2008 R2. Both 32- and 64-bit editions are supported.
Microsoft Visual C++ 2010 runtime.
A cluster with a Hadoop node running the Hive server. See Starting the Hive server.
To install the DataStax ODBC driver on a Windows platform:
Procedure
1. Download the driver from Client Libraries and CQL Drivers.
2. Double-click the downloaded file and follow the wizard's instructions.
Configuring the driver
Set up the DataStax Hive ODBC driver for access by your BI tool.
About this task
Set up the DataStax ODBC driver for access by your BI tool.
Procedure
1. Click Start Program Files > DataStax Hive ODBC Connector > ODBC Driver Manager.
2. Click the Drivers tab to verify that the driver is present.
3. Create either a User or System DSN (data source name) for your BI tool connection.
a) Click the User DSN or System DSN tab.
b) Click Add > DataStax Hive ODBC Connector > Finish.
c) In DataStax Hive ODBC Connector Setup, enter the following:
81
DSE Analytics with Hadoop
Data Source
Name
The name for your DSN.
Description
Optional.
Host
IP or hostname of your Hive server.
Port
Listening port for the Hive service.
Database
By default, all tables reside within the default database. To check for the
appropriate database, use the show databases Hive command.
d) Click Test.
The test results are displayed.
Note: If your DataStax Enterprise cluster is on Amazon EC2, you must open the listing port
for the Hive Server. For more information, refer to Installing a DSE cluster on EC2.
4. To configure the advanced options, see Appendix C in the DataStax Hive ODBC Connector User Guide
for Windows:
Start > Program Files > DataStax Hive ODBC Connector > User's Guide
Using the DataStax ODBC driver for Hive
After configuring the ODBC data source for Hive, connect and pull data from Hive using any compliant BI
tool.
About this task
After configuring the ODBC data source for Hive, you can connect and pull data from Hive using any
compliant BI tool. For example, to retrieve data using Microsoft Excel:
Procedure
1. Use the data connection wizard to select your new ODBC data source:
2. In Connect to OBDC Data Source, select DSE2 Hive > Next.
3. Select one or more data objects (or construct a query) to retrieve the data, and then click Finish.
82
DSE Analytics with Hadoop
Results
After the ODBC query is executed and the data is retrieved, a Hive MapReduce job runs on the server:
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks is set to 0 since there's no reduce operator
Starting Job = job_201208230939_0006,
Tracking URL = http://localhost:50030/jobdetails.jsp?
jobid=job_201208230939_0006
Kill Command = ./dse hadoop job
-Dmapred.job.tracker=127.0.0.1:8012 -kill job_201208230939_0006
Hadoop job information for Stage-1: number of mappers: 1; number of reducers:
0
2012-08-23 12:44:39,795 Stage-1 map = 0%, reduce = 0%
2012-08-23 12:44:42,824 Stage-1 map = 100%, reduce = 0%
2012-08-23 12:44:44,833 Stage-1 map = 100%, reduce = 100%
Ended Job = job_201208230939_0006
MapReduce Jobs Launched:
Job 0: Map: 1
HDFS Read: 0 HDFS Write: 0 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
Using Mahout
DataStax Enterprise integrates Apache Mahout, a Hadoop component that offers machine learning
libraries.
About this task
DataStax Enterprise integrates Apache Mahout, a Hadoop component that offers machine learning
libraries. Mahout facilitates building intelligent applications that learn from data and user input. Machine
learning use cases are many and some, such as the capability of web sites to recommend products to
visitors based on previous visits, are notorious.
Currently, Mahout jobs that use Lucene features are not supported.
Running the Mahout demo
The DataStax Enterprise installation includes a Mahout demo. The demo determines with some
percentage of certainty which entries in the input data remained statistically in control and which have not.
The input data is time series historical data. Using the Mahout algorithms, the demo classifies the data
into categories based on whether it exhibited relatively stable behavior over a period of time. The demo
produces a file of classified results. This procedure describes how to run the Mahout demo.
Procedure
1. After installing DataStax Enterprise, start an analytics node.
83
DSE Analytics with Hadoop
2. Go to the demos directory in one of the following locations:
• Tarball installs: install_location/demos/mahout
• Packaged installs: /usr/share/dse-demos/mahout
3. Run the script in the demos directory. For example, on Linux:
./run_mahout_example.sh
If you are running OpsCenter, you can now view the Hadoop job progress:
When the demo completes, a message appears on the standard output about the location of the output
file. For example:
The output is in /tmp/clusteranalyze.txt
Using Mahout commands in DataStax Enterprise
Run Mahout commands on the dse command line.
About this task
You can run Mahout commands on the dse command line. For example on Mac OS X, to get a list of which
commands are available:
$ cd install_directory
$ bin/dse mahout
The list of commands appears.
Mahout command line help
You use one of these commands as the first argument plus the help option:
$ cd install_directory
$ bin/dse mahout arff.vector --help
The output is help on the arff.vector command.
Add Mahout classes to the class path, execute Hadoop command
You use Hadoop shell commands to work with Mahout. Using this syntax first adds Mahout classes to the
class path, and then executes the Hadoop command:
$ cd install_directory
$ bin/dse mahout hadoop fs -text mahout file | more
The Apache web site offers an in-depth tutorial.
Using Pig
DataStax Enterprise includes a Cassandra File System (CFS) enabled Apache Pig Client to provide a highlevel programming environment for MapReduce coding.
About this task
DataStax Enterprise (DSE) includes a CassandraFS enabled Apache Pig Client. Pig is a higher-level
programming environment for MapReduce coding. You can explore big data sets using the Pig Latin data
84
DSE Analytics with Hadoop
flow language for programmers. Relations, which are similar to tables, are constructed of tuples, which
correspond to the rows in a table. Unlike a relational database table, Pig relations do not require every
tuple to contain the same number of fields. Fields in the same position (column) need not be of the same
type. Using Pig, you can devise logic for data transformations, such as filtering data and grouping relations.
The transformations occur during the MapReduce phase.
Configure the job tracker node for the node running Pig as you would for any analytics (Hadoop) node. Use
the dsetool commands to manage the job tracker. After configuration, Pig clients automatically select the
correct job tracker node on startup. Pig programs are compiled into MapReduce jobs, executed in parallel
by Hadoop, and run in a distributed fashion on a local or remote cluster.
Support for CQL collections
Pig in DataStax Enterprise supports CQL collections. Pig-supported types must be used.
Running the Pig demo
Examples demonstrate how to use Pig to work with CQL tables.
About this task
Three examples demonstrate how to use Pig to work with CQL tables.
•
How to save Pig relations from/to Cassandra
•
Pig uses a single tuple.
How to work with a Cassandra compound primary key in Pig
•
Pig uses three tuples, one for the partition key and two for the two clustering columns.
How to use Pig to set up logic for exploring library data
This example from the Cassandra and Pig tutorial shows how to copy public library data into
Cassandra, add logic to save the data to a Pig relation, execute programs by running MapReduce jobs,
and view results in a Cassandra table.
Start Pig
Procedure
1. Start DataStax Enterprise as an analytics (Hadoop) node:
•
Packaged installs:
Set HADOOP_ENABLED=1 in the /etc/default/dse file, and then use this command to start an
analytics node:
•
$ sudo service dse start
Tarball installs:
$ DSE_install_directory/bin/dse cassandra -t
2. Start the Pig shell:
•
•
Packaged installs: $ dse pig
Tarball installs: $ DSE_install_directory/bin/dse pig
The Pig grunt prompt appears, and you can now enter Pig commands.
Example: Save Pig relations from/to Cassandra
How to merge the data from two CQL tables having simple primary keys using Pig.
85
DSE Analytics with Hadoop
About this task
For Pig to access data in Cassandra, the target keyspace and table must already exist. Pig can save data
from a Pig relation to a table in Cassandra and from a Cassandra table to a pig relation, but it cannot
create the table. This example shows how to merge the data from two CQL tables having simple primary
keys using Pig.
A subsequent example shows how to merge data from CQL tables having compound primary keys into one
CQL table using Pig.
Procedure
1. Start cqlsh.
2. Using cqlsh, create and use a keyspace named, for example, cql3ks.
cqlsh> CREATE KEYSPACE cql3ks WITH replication =
{'class': 'SimpleStrategy', 'replication_factor': 1 };
cqlsh> USE cql3ks;
3. Create a two-column (a and b) Cassandra table named simple_table1 and another two-column (x and
y) table named simple_table2. Insert data into the tables.
cqlsh:cql3ks> CREATE TABLE simple_table1 (a int PRIMARY KEY, b int);
cqlsh:cql3ks> CREATE TABLE simple_table2 (x int PRIMARY
KEY, y int);
cqlsh:cql3ks> INSERT INTO simple_table1 (a,b) VALUES
(1,1);
cqlsh:cql3ks> INSERT INTO simple_table1 (a,b) VALUES
(2,2);
cqlsh:cql3ks> INSERT INTO simple_table1 (a,b) VALUES
(3,3);
cqlsh:cql3ks> INSERT INTO simple_table2 (x, y) VALUES
(4,4);
cqlsh:cql3ks> INSERT INTO simple_table2 (x, y) VALUES
(5,5);
cqlsh:cql3ks> INSERT INTO simple_table2 (x, y) VALUES
(6,6);
4. Using Pig, add logic to load the data (4, 5, 6) from the Cassandra simple_table2 table into a Pig
relation. In DataStax Enterprise 4.0.4 and later, use USING CqlNativeStorage instead of USING
CqlStorage.
grunt> moretestvalues= LOAD 'cql://cql3ks/simple_table2/' USING CqlStorage;
5. Convert the simple_table2 table data to a tuple. The key column is a chararray, 'a'.
grunt> insertformat= FOREACH moretestvalues GENERATE
TOTUPLE(TOTUPLE('a',x)),TOTUPLE(y);
86
DSE Analytics with Hadoop
6. Save the relation to the Cassandra simple_table1 table. In DataStax Enterprise 4.0.4 and later, use
USING CqlNativeStorage instead of USING CqlStorage.
grunt> STORE insertformat INTO
'cql://cql3ks/simple_table1?output_query=UPDATE
+cql3ks.simple_table1+set+b+%3D+%3F'
USING CqlStorage;
Pig uses a URL-encoded prepared statement to store the relation to Cassandra. The cql:// URL is
followed by an output_query, which specifies which key should be used in the command. The rest of the
arguments, the "?"s, for the prepared statement are filled in by the values related to that key in Pig.
7. On the cqlsh command line, check that the simple_table1 table now contains its original values plus the
values from the simple_table2 table:
cqlsh:cql3ks> SELECT * FROM simple_table1;
a | b
--+-5 | 5
1 | 1
2 | 2
4 | 4
6 | 6
3 | 3
Example: Handle a compound primary key
Work with CQL tables in Pig. The tables use compound primary keys. You create the tables in cqlsh and
merge them using Pig.
About this task
This example, like the previous one, shows you how to work with CQL tables in Pig. The previous example
used tables having a simple primary key. The tables in this example use compound primary keys. You
create the tables in cqlsh and merge them using Pig.
87
DSE Analytics with Hadoop
88
DSE Analytics with Hadoop
Procedure
1. Create a four-column (a, b, c, d) Cassandra table named table1 and another five-column (id, x, y, z,
data) table named table2.
cqlsh:cql3ks> CREATE TABLE table1 (
a int,
b int,
c text,
d text,
PRIMARY KEY (a,b,c)
);
cqlsh:cql3ks> CREATE TABLE table2 (
id int PRIMARY KEY,
x int,
y int,
z text,
data text
);
2. Insert data into the tables.
cqlsh:cql3ks> INSERT INTO table1 (a, b , c , d )
VALUES ( 1,1,'One','match');
cqlsh:cql3ks> INSERT INTO table1 (a, b , c , d )
VALUES ( 2,2,'Two','match');
cqlsh:cql3ks> INSERT INTO table1 (a, b , c , d )
VALUES ( 3,3,'Three','match');
cqlsh:cql3ks> INSERT INTO table1 (a, b , c , d )
VALUES ( 4,4,'Four','match');
cqlsh:cql3ks> INSERT INTO table2 (id, x, y, z,data)
VALUES (1,5,6,'Fix','nomatch');
cqlsh:cql3ks> INSERT INTO table2 (id, x, y, z,data)
VALUES (2,6,5,'Sive','nomatch');
cqlsh:cql3ks> INSERT INTO table2 (id, x, y, z,data)
VALUES (3,7,7,'Seven','match');
cqlsh:cql3ks> INSERT INTO table2 (id, x, y, z,data)
VALUES (4,8,8,'Eight','match');
cqlsh:cql3ks> INSERT INTO table2 (id, x, y, z,data)
VALUES (5,9,10,'Ninen','nomatch');
3. Using Pig, add logic to load the data from the Cassandra table2 to a Pig relation. In DataStax Enterprise
4.0.4 and later, use USING CqlNativeStorage instead of USING CqlStorage.
grunt> moredata = load 'cql://cql3ks/table2' USING CqlStorage;
4. Convert the data to a tuple.
grunt> insertformat = FOREACH moredata GENERATE TOTUPLE
(TOTUPLE('a',x),TOTUPLE('b',y),
TOTUPLE('c',z)),TOTUPLE(data);
During the actual data processing, the data is formatted as follows:
((PartitionKey_Name,Value),(ClusteringKey_1_name,Value)...)
(ArgValue1,ArgValue2,ArgValue3,...)
5. Save the Pig relation to the Cassandra table1 table. The data from table 1 and table 2 will be merged.
grunt> STORE insertformat INTO 'cql://cql3ks/table1?output_query=UPDATE
%20cql3ks.table1%20SET%20d%20%3D%20%3F' USING CqlStorage;
The cql:// URL includes a prepared statement, described later, that needs to be copied/pasted as a
continuous string (no spaces or line breaks).
6. In cqlsh, query table1 to check that the data from table1 and table2 have been merged.
89
DSE Analytics with Hadoop
cqlsh:cql3ks> SELECT * FROM table1;
a | b | c
| d
---+----+-------+--------5 | 6 |
Fix | nomatch
1 | 1 |
One |
match
8 | 8 | Eight |
match
2 | 2 |
Two |
match
4 | 4 | Four |
match
7 | 7 | Seven |
match
6 | 5 | Sive | nomatch
9 | 10 | Ninen | nomatch
3 | 3 | Three |
match
Example: Explore library data
Install library data that is encoded in UTF-8 format and use a pig script.
About this task
This example uses library data from the Institute of Library and Museum Services, encoded in UTF-8
format. Download the formatted data for this example now.
DataStax Enterprise installs files in the following directory that you can use to run through this example
using a pig script instead of running Pig commands manually.
•
•
Packaged installs: /usr/share/dse-demos/pig/cql
Tarball installs: install-location/demos/pig/cql
Using the files is optional. To use the files, copy/paste the commands in steps 2-3 from the librarypopulate-cql.txt file and execute steps 7-10 automatically by running the library-cql.pig script.
Procedure
1. Unzip libdata.csv.zip and give yourself permission to access the downloaded file. On the Linux
command line, for example:
$ chmod 777 libdata.csv
2. Create and use a keyspace called libdata.
cqlsh:libdata> CREATE KEYSPACE libdata WITH replication =
{'class': 'SimpleStrategy', 'replication_factor': 1 };
cqlsh:libdata> USE libdata;
3. Create a table for the library data that you downloaded.
cqlsh:libdata> CREATE TABLE libout ("STABR" TEXT, "FSCSKEY" TEXT, "FSCS_SEQ"
TEXT,
"LIBID" TEXT, "LIBNAME" TEXT, "ADDRESS" TEXT, "CITY" TEXT,
"ZIP" TEXT, "ZIP4" TEXT, "CNTY" TEXT, "PHONE" TEXT,
"C_OUT_TY" TEXT,
"C_MSA" TEXT, "SQ_FEET" INT, "F_SQ_FT" TEXT, "L_NUM_BM"
INT,
"F_BKMOB" TEXT, "HOURS" INT, "F_HOURS" TEXT, "WKS_OPEN"
INT,
"F_WKSOPN" TEXT, "YR_SUB" INT, "STATSTRU" INT, "STATNAME"
INT,
"STATADDR" INT, "LONGITUD" FLOAT, "LATITUDE" FLOAT,
"FIPSST" INT,
"FIPSCO" INT, "FIPSPLAC" INT, "CNTYPOP" INT, "LOCALE" TEXT,
"CENTRACT" FLOAT, "CENBLOCK" INT, "CDCODE" TEXT, "MAT_CENT"
TEXT,
90
DSE Analytics with Hadoop
"MAT_TYPE" INT, "CBSA" INT, "MICROF" TEXT,
PRIMARY KEY ("FSCSKEY", "FSCS_SEQ"));
4. Import data into the libout table from the libdata.csv file that you downloaded.
cqlsh:libdata> COPY libout ("STABR","FSCSKEY","FSCS_SEQ","LIBID","LIBNAME",
"ADDRESS","CITY","ZIP","ZIP4","CNTY","PHONE","C_OUT_TY",
"C_MSA","SQ_FEET","F_SQ_FT","L_NUM_BM","F_BKMOB","HOURS",
"F_HOURS","WKS_OPEN","F_WKSOPN","YR_SUB","STATSTRU","STATNAME",
"STATADDR","LONGITUD","LATITUDE","FIPSST","FIPSCO","FIPSPLAC",
"CNTYPOP","LOCALE","CENTRACT","CENBLOCK","CDCODE","MAT_CENT",
"MAT_TYPE","CBSA","MICROF") FROM 'libdata.csv' WITH
HEADER=TRUE;
In the FROM clause of the COPY command, use the path to libdata.csv in your environment.
5. Check that the libout table contains the data you copied from the downloaded file.
cqlsh:libdata> SELECT count(*) FROM libdata.libout LIMIT 20000;
count
------17598
6. Create a table to hold results of Pig relations.
cqlsh:libdata> CREATE TABLE libsqft (
year INT,
state TEXT,
sqft BIGINT,
PRIMARY KEY (year, state)
);
7. Using Pig, add a plan to load the data from the Cassandra libout table to a Pig relation. In DataStax
Enterprise 4.0.4 and later, use USING CqlNativeStorage instead of USING CqlStorage.
grunt> libdata = LOAD 'cql://libdata/libout' USING CqlStorage();
8. Add logic to remove data about outlet types other than books-by-mail (BM). The C_OUT_TY column
uses BM and other abbreviations to identify these library outlet types:
•
•
•
•
CE–Central Library
BR–Branch Library
BS–Bookmobile(s)
BM–Books-by-Mail Only
grunt> book_by_mail = FILTER libdata BY C_OUT_TY == 'BM';
grunt> DUMP book_by_mail;
9. Add logic to filter out the library data that has missing building size data, define the schema for
libdata_buildings, and group data by state. The STABR column contains the state codes. GROUP
creates the state_grouped relation. Pig gives the grouping field the default alias group. Process each
row to generate a derived set of rows that aggregate the square footage of each state group.
grunt> libdata_buildings = FILTER libdata BY SQ_FEET > 0;
grunt> state_flat = FOREACH libdata_buildings GENERATE STABR AS
State,SQ_FEET AS SquareFeet;
grunt> state_grouped = GROUP state_flat BY State;
grunt> state_footage = FOREACH state_grouped GENERATE
group as State,SUM(state_flat.SquareFeet)
AS TotalFeet:int;
grunt> DUMP state_footage;
The MapReduce job completes successfully and the output shows the square footage of the buildings.
. . .
91
DSE Analytics with Hadoop
(UT,1510353)
(VA,4192931)
(VI,31875)
(VT,722629)
(WA,3424639)
(WI,5661236)
(WV,1075356)
(WY,724821)
10.Add logic to filter the data by year, state, and building size, and save the relation to Cassandra using
the cql:// URL. In DataStax Enterprise 4.0.4 and later, use USING CqlNativeStorage instead of USING
CqlStorage. The URL includes a prepared statement, described later.
grunt> insert_format= FOREACH state_footage GENERATE
TOTUPLE(TOTUPLE('year',2011),TOTUPLE('state',State)),TOTUPLE(TotalFeet);
grunt> STORE insert_format INTO 'cql://libdata/libsqft?output_query=UPDATE
%20libdata.
libsqft%20SET%20sqft%20%3D%20%3F' USING CqlStorage;
When the MapReduce job completes, a message appears that the records were written successfully.
11.In CQL, query the libsqft table to see the Pig results now stored in Cassandra.
cqlsh> SELECT * FROM libdata.libsqft;
year | state | sqft
------+-------+---------2011 |
AK |
570178
2011 |
AL | 2792246
. . .
2011 |
2011 |
WV |
WY |
1075356
724821
Data access using storage handlers
To execute Pig programs directly on data that is stored in Cassandra, use one of the DataStax Enterprise
storage handlers.
The DataStax Enterprise Pig driver uses the Cassandra File System (CassandraFS) instead of the Hadoop
distributed file system (HDFS). Apache Cassandra, on the other hand, includes a Pig driver that uses the
Hadoop Distributed File System (HDFS).
To execute Pig programs directly on data stored in Cassandra, you use one of the DataStax Enterprise
storage handlers:
Table Format
Storage Handler
URL
CQL
CqlNativeStorage()
cql:// DataStax 4.0.4 and later
CQL
CqlStorage()
cql://
storage engine
CassandraStorage()
cassandra://
The CqlStorage handler is deprecated and slated for removal at some point in the future. Use
the CqlNativeStorage handler and the cql:// URL for new pig applications. Migrate all tables to
CqlNativeStorage as soon as possible in preparation for the removal of the CqlStorage handler.
Migrating compact tables with clustering columns to CqlNativeStorage format
The CqlNativeStorage handler uses native paging through the DataStax java driver to communicate with
the underlying Cassandra cluster. In DataStax Enterprise 4.0.4, to use applications having compact tables
92
DSE Analytics with Hadoop
with clustering columns in the CqlStorage format, you need to migrate tables to the CqlNativeStorage
format. Attempting to run Pig commands on compact tables in the CqlStorage format results in an
exception. You can, however, run Pig commands on non-compact tables in the CqlStorage format.
To migrate tables from CqlStorage to CqlNativeStorage format:
1. Identify Pig functions that interact with compact tables in CqlStorage format. For example, suppose
you identify a command that adds logic to load the data to a Pig relation from the compact table tab in
keyspace ks.
x = LOAD 'cql://ks/tab' USING CqlStorage();
2. Change CqlStorage() to USING CqlNativeStorage().
-- Old function
x = LOAD 'cql://ks/tab' USING CqlNativeStorage(); -- New function
URL format for CqlNativeStorage
The url format for CqlNative Storage is:
cql://[username:[email protected]]<keyspace>/<table>[?
[page_size=<size>]
[&columns=<col1,col2>]
[&output_query=<prepared_statement_query>]
[&cql_input=<prepared_statement_query>]
[&where_clause=<clause>]
[&split_size=<size>]
[&partitioner=<partitioner>]
[&use_secondary=true|false]]
[&init_address=<host>]
[&native_port=<port>]]
Where:
•
•
•
•
•
•
•
•
•
•
page_size -- the number of rows per page
columns -- the select columns of CQL query
output_query -- the CQL query for writing in a prepared statement format
input_cql -- the CQL query for reading in a prepared statement format
where_clause -- the where clause on the index columns, which needs url encoding
split_size -- number of rows per split
partitioner -- Cassandra partitioner
use_secondary -- to enable pig filter partition push down
init_address -- the IP address of the target node
native_port -- the listen address of the target node
URL format for CqlStorage
The url format for CqlStorage is:
cql://[username:[email protected]]<keyspace>/<table>[?
[page_size=<size>]
[&columns=<col1,col2>]
[&output_query=<prepared_statement_query>]
[&where_clause=<clause>]
[&split_size=<size>]
[&partitioner=<partitioner>]
[&use_secondary=true|false]]
[&init_address=<host>]
[&rpc_port=<port>]]
Where:
•
•
•
page_size -- the number of rows per page
columns -- the select columns of CQL query
output_query -- the CQL query for writing in a prepared statement format
93
DSE Analytics with Hadoop
•
•
•
•
•
•
where_clause -- the where clause on the index columns, which needs url encoding
split_size -- number of rows per split
partitioner -- Cassandra partitioner
use_secondary -- to enable pig filter partition push down
init_address -- the IP address of the target node
rpc_port -- the listen address of the target node
Working with legacy Cassandra tables
Use the CassandraStorage() handler and cfs:// url to work with Cassandra tables that are in the storage
engine (CLI/Thrift) format in Pig. Legacy tables are created using Thrift, CLI, or using the WITH COMPACT
STORAGE directive in CQL. Thrift applications require that you configure Cassandra for connection to your
application using the rpc connections instead of the default native_transport connection.
URL format for CassandraStorage
The URL format for CassandraStorage is:
cassandra://[username:[email protected]]<keyspace>/<columnfamily>[?
slice_start=<start>&slice_end=<end>
[&reversed=true]
[&limit=1]
[&allow_deletes=true]
[&widerows=true]
[&use_secondary=true]
[&comparator=<comparator>]
[&split_size=<size>]
[&partitioner=<partitioner>]
[&init_address=<host>]
[&rpc_port=<port>]]
CQL data access
Use the CqlNativeStorage handler with the input_cql statement or use the output_query statement that was
available in earlier releases.
In DataStax Enterprise 4.0.4, to access data in CQL tables, use the CqlNativeStorage handler with the new
input_cql statement or use the output_query statement that was available in earlier releases.
In DataStax Enterprise 4.0-4.0.3, to access data in CQL tables, use the CqlStorage() handler. To access
data in the CassandraFS, the target keyspace and table must already exist. Data in a Pig relation can be
stored in a Cassandra table, but Pig will not create the table.
The Pig LOAD function pulls Cassandra data into a Pig relation through the storage handler as shown in
these examples:
•
DataStax Enterprise 4.0.4
•
<pig_relation_name> = LOAD 'cql://<keyspace>/<table>'
USING CqlNativeStorage(); -- DataStax Enterprise 4.0.4
DataStax Enterprise 4.0 - 4.0.3
<pig_relation_name> = LOAD 'cql://<keyspace>/<table>'
USING CqlStorage(); -- DataStax Enterprise 4.0 - 4.0.3
DataStax Enterprise supports these Pig data types:
•
•
•
•
•
•
94
int
long
float
double
boolean
chararray
DSE Analytics with Hadoop
The Pig demo examples include using the LOAD command.
LOAD schema
The LOAD Schema is:
(colname:colvalue, colname:colvalue, … )
where each colvalue is referenced by the Cassandra column name.
Accessing data using input_cql and CqlNativeStorage handler
The input_cql statement contains the following components:
•
•
•
A SELECT statement that includes the partition key columns
A WHERE clause that includes the range of the columns consistent with the order in the cluster and in
the following format:
WHERE token(partitionkey) > ? and token(partitionkey) <?
The value of the native_port
For example, the input_cql statement before encoding might look like this:
'SELECT * FROM ks.tab where token(key) > ? and token (key) <= ?' USING
CqlNativeStorage();
Append the encoded statement as an argument to the pig Load command using the ?input_cql= syntax.
x = LOAD 'cql://ks/tab?input_cql=SELECT%20*%20FROM%20ks.tab%20where
%20token(key)%20%3E%20%3F%20and%20token%20(key)%20%3C%3D%20%3F' USING
CqlNativeStorage();
Use an ampersand to append additional parameters. For example, to modify the port used by the Java
Driver, append the following parameter and port number.
&native_port=9042
The entire migrated Pig command would look like this:
x = LOAD 'cql://ks/tab?input_cql=SELECT%20*%20FROM%20ks.tab
%20where%20token(key)%20%3E%20%3F%20and%20token%20(key)%20%3C%3D
%20%3F&amp;native_port=9042' USING CqlNativeStorage();
Optional input_cql parameters
You can use the following list of parameters with input_cql in DataStax Enterprise 4.0.4 and later as shown
by the example in the last section. The ampersand must preface the parameter.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
&native_port=<native_port>
&core_conns=<core_conns>
&max_conns=<max_conns>
&min_simult_reqs=<min_simult_reqs>
&max_simult_reqs=<max_simult_reqs>
&native_timeout=<native_timeout>
&native_read_timeout=<native_read_timeout>
&rec_buff_size=<rec_buff_size>
&send_buff_size=<send_buff_size>
&solinger=<solinger>
&tcp_nodelay=<tcp_nodelay>
&reuse_address=<reuse_address>
&keep_alive=<keep_alive>
&auth_provider=<auth_provider>
&trust_store_path=<trust_store_path>
&key_store_path=<key_store_path>
95
DSE Analytics with Hadoop
•
•
•
•
&trust_store_password=<trust_store_password>
&key_store_password=<key_store_password>
&cipher_suites=<cipher_suites>
&input_cql=<input_cql>
Handling special characters in the CQL
If the input_cql or output_query to a Pig function contains special characters, you need to url-encode a
prepared statement to make special characters readable by Pig.
CQL pushdown filter
Optimize the processing of the data by moving filtering expressions in Pig as close to the data source as
possible.
DataStax Enterprise includes a CqlStorage URL option, use_secondary. Setting the option to true
optimizes the processing of the data by moving filtering expressions in Pig as close to the data source as
possible. To use this capability:
•
•
Create an index for the Cassandra table.
For Pig pushdown filtering, the index must have the same name as the column being indexed.
Include the use_secondary option with a value of true in the url format for the storage handler. The
option name reflects the term that used to be used for a Cassandra index: secondary index. For
example:
newdata = LOAD 'cql://ks/cf_300000_keys_50_cols?use_secondary=true'
USING CqlStorage();
-- DataStax Enterprise 4.0 4.0.3
newdata = LOAD 'cql://ks/cf_300000_keys_50_cols?use_secondary=true'
USING CqlNativeStorage();
-- DataStax Enterprise
4.0.4
Saving a Pig relation to Cassandra
The Pig STORE command pushes data from a Pig relation to Cassandra through the CqlNativeStorage
handler.
The Pig STORE command pushes data from a Pig relation to Cassandra through the CqlStorage handler
(DataStax Enterprise 4.0 - 4.0.3) or CqlNativeStorage handler (DataStax Enterprise 4.0.4 and later):
STORE <relation_name> INTO 'cql://<keyspace>/<column_family>?<prepared
statement>'
USING CqlStorage();
-- DataStax Enterprise 4.0 - 4.0.3
STORE <relation_name> INTO 'cql://<keyspace>/<column_family>?<prepared
statement>'
USING CqlNativeStorage();
-- DataStax Enterprise 4.0.4
Store schema
The input schema for Store is:
(value, value, value)
where each value schema has the name of the column and value of the column value.
The output schema for Store is:
(((name, value), (name, value)), (value ... value), (value ... value))
where the first tuple is the map of partition key and clustering columns. The rest of the tuples are the list of
bound values for the output in a prepared CQL query.
96
DSE Analytics with Hadoop
Creating a URL-encoded prepared statement
Pig demo examples set up a prepared CQL query using the output_query statement.
About this task
The Pig demo examples show the steps required for setting up a prepared CQL query using the
output_query statement available in DataStax Enterprise 4.0 - 4.0.3:
Procedure
1. Format the data
The example of saving Pig relations from/to Cassandra shows the output schema: the name of the
simple_table1 table primary key 'a', represented as a chararray in the relation is paired with a value in
the simple_table2 table. In this case, the key for simple_table1 table is only a partitioning key, and only
a single tuple is needed.
The Pig statement to add (moredata) fields to a tuple is:
grunt> insertformat= FOREACH morevalues GENERATE
TOTUPLE(TOTUPLE('a',x)),TOTUPLE(y);
The example of exploring library data works with more complicated data, a partition key and clustering
column:
grunt> insertformat = FOREACH moredata GENERATE
TOTUPLE(TOTUPLE('a',x),TOTUPLE('b',y),TOTUPLE('c',z)),TOTUPLE(data);
2. Construct the prepared query
The output query portion of the cql:// URL is the prepared statement. The prepared statement must be
url-encoded to make special characters readable by Pig.
The example of saving Pig relations from/to Cassandra shows how to construct a prepared query:
'cql://cql3ks/simple_table1?output_query=UPDATE+cql3ks.simple_table1+set+b+
%3D+%3F'
The key values of the simple_table1 table are automatically transformed into the 'WHERE (key) ='
clause to form the output_query portion of a prepared statement.
3. Execute the query
To update the simple_table1 table using the values in the simple_table2 (4-6), the prepared statement
is executed using these WHERE clauses when the MapReduce job runs:
... WHERE a = 5
... WHERE a = 4
... WHERE a = 6
This output_query in Pig statement forms the '...' url-encoded portion of the prepared statement:
grunt> STORE insertformat INTO
'cql://cql3ks/simple_table1?output_query=UPDATE
+cql3ks.simple_table1+set+b+%3D+%3F'
USING CqlStorage;
Decoded the UPDATE statement is:
UPDATE cql3ks.simple_table1 SET b = ?
The prepared statement represents these queries:
UPDATE cql3ks.test SET b = 5 WHERE a = 5;
UPDATE cql3ks.test set b = 4 WHERE a = 4;
UPDATE cql3ks.test set b = 6 WHERE a = 6;
97
DSE Analytics with Hadoop
Sqoop
Migrating data using Sqoop topics.
About Sqoop
You can use Sqoop for migrating data and supports password authentication for Sqoop operations.
Sqoop is an Apache Software Foundation tool for transferring data between an RDBMS data source and
Hadoop or between other data sources, such as NoSQL.
DataStax Enterprise support for Sqoop empowers you to import data from an external data source to the
Cassandra File System (CassandraFS), the Hadoop File System (HDFS) counterpart. An analytics node
runs the MapReduce job that imports data from a data source using Sqoop.
You need a JDBC driver for the RDBMS or other type of data source.
Importing data
You can import data from any JDBC-compliant data source. For example:
•
•
•
•
•
DB2
MySQL
Oracle
SQL Server
Sybase
Running the Sqoop demo
Run SQL commands to put the data from a CSV file into a MySQL table in a MySQL database. Then
import the SQL data from MySQL to a CQL table in Cassandra.
About this task
The Sqoop demo migrates the data from a MySQL table to text files in the CassandraFS. The Sqoop data
migration demo uses the MySQL database and data from the North American Numbering Plan. This data
consists of the area-code (NPA) and telephone number (Nxx) for the USA and Canada.
Before you begin
To run the demo, you need:
•
•
•
•
•
•
•
•
Latest version of Oracle Java SE Development Kit (JDK) 7. The JRE alone will not work.
An installation of MySQL
Sufficient MySQL database privileges to create database objects
A JDBC driver for MySQL in the sqoop/lib directory
The connection string that is appropriate for the JDBC driver
One or more DSE nodes running the Analytics workload to run the Hadoop job that actually imports
data from the external data source
A PATH environment variable that includes the bin directory of the DSE installation
To run this demo and import data to nodes in a cluster, the database permissions must be granted to
the nodes. For example, use the GRANT ALL command to grant MySQL access to the hosts.
Procedure
To run the Sqoop demo on a single node on a Mac, for example, follow these steps:
1. Install MySQL and download the JDBC driver for MySQL from the MySQL site. This example uses
mysql-connector-java-5.1.29-bin.jar.
98
DSE Analytics with Hadoop
2. Copy the JDBC driver for MySQL to the sqoop/lib directory.
• Packaged installs: install_location/usr/share/dse/sqoop/lib
• Tarball installs: /resources/sqoop/lib
3. On the operating system command line, start the MySQL daemon. For example:
$ sudo ./mysqld_safe --user=mysql
4. Start MySQL and create the demo database:
mysql> CREATE DATABASE npa_nxx_demo ;
5. Then connect to the database and create the table:
mysql> USE npa_nxx_demo;
mysql> CREATE TABLE npa_nxx (
npa_nxx_key varchar(16) NOT NULL,
npa
varchar(3) DEFAULT NULL,
nxx
varchar(3) DEFAULT NULL,
lat
varchar(8) DEFAULT NULL,
lon
varchar(8) DEFAULT NULL,
linetype
varchar(1) DEFAULT NULL,
state
varchar(2) DEFAULT NULL,
city
varchar(36) DEFAULT NULL,
PRIMARY KEY (npa_nxx_key)
) ENGINE=InnoDB DEFAULT CHARSET=latin1;
6. Locate the demos/sqoop directory:
• Packaged installs: /usr/share/dse-demos/sqoop
• Tarball installs: install_location/demos/sqoop
7. Populate the table by loading the CSV file in the demos/sqoop directory.
mysql> LOAD DATA LOCAL INFILE 'npa_nxx.csv'
INTO TABLE npa_nxx_demo.npa_nxx
FIELDS TERMINATED BY ','
ENCLOSED BY '"'
LINES TERMINATED BY '\n';
MySQL returns the following message:
Query OK, 105291 rows affected (1.01 sec) Records: 105291 Deleted: 0
Skipped: 0 Warnings: 0
8. Start DSE as an analytics node.
•
Packaged installs:
•
Edit /etc/default/dse, set HADOOP_ENABLED=1 in the cassandra.yaml to start the DSE
service.
Tarball installs:
From the installation directory of the DSE installation directory, use the -t option to start an analytics/
Hadoop node:
bin/dse cassandra -t
9. Use the dse command in the bin directory to migrate the data from the MySQL table to text files in the
directory npa_nxx in the CassandraFS.
Use the database username and password or -P instead of --password to be prompted for the database
password. If the database account is not password-protected, just omit the password option.
$ bin/dse sqoop import --connect
jdbc:mysql://127.0.0.1/npa_nxx_demo
--username mysql
--password <password>
--table npa_nxx
99
DSE Analytics with Hadoop
--target-dir /npa_nxx
DSE returns this message: INFO mapreduce.ImportJobBase: Retrieved 105291 records.
Validating import results in a cluster
View the results of an import in the Cassandra File System.
Use this command to view the results in the Cassandra File System:
./dse hadoop fs -ls /npa_nxx
Depending on the number of DataStax Enterprise analytics nodes and task tracker configuration, the
output shows a number of files in the directory, part-m-0000n, where 'n' ranges from 0 to the number of
tasks that were executed as part of the Hadoop job.
The contents of these files can be viewed using this command:
./dse hadoop fs -cat /npa_nxx/part-m-00000
By varying the number of tasks (the 00000), the output looks something like this:
361991,361,991,27.73,097.40,L,TX,Corpus Christi
361992,361,992,27.73,097.40,L,TX,Corpus Christi
361993,361,993,27.73,097.40,L,TX,Corpus Christi
361994,361,994,27.73,097.40,L,TX,Corpus Christi
361998,361,998,27.79,097.90,L,TX,Agua Dulce
361999,361,999,27.80,097.40,W,TX,Padre Island National Seashore
As shown in the output, the CSV file format that Sqoop requires does not include optional spaces in the
delimiter.
Migrating data to a Cassandra table
Using Hive to load the imported data into a Cassandra CQL table.
After importing data into text files in the CassandraFS, you can use Hive to load the imported data into a
Cassandra CQL table. In Hive, create two external tables, one to map the CassandraFS data into Hive and
the other to map and store the data into Cassandra. Download sample commands for accomplishing this
task. Adapt the sample commands to your environment. For example, in the npa_from_cfs table, change
the path in the AS TEXTFILE LOCATION clause to match the path to the data in the CassandraFS.
Using the Sqoop import command with Cassandra options, you can import data into a legacy Cassandra
table and access the table using Thrift. In the previous release, you could access the legacy Cassandra
table using cqlsh. In DataStax Enterprise 4.0 and later, cqlsh support for legacy tables has been removed.
About the generated Sqoop JAR file
After running the dse sqoop import command, a Java class appears in the DSE installation bin directory.
After running the dse sqoop import command, a Java class appears in the DSE installation bin directory.
This file, by default named npa_nxx.java after the demo table, can encapsulate one row of the imported
data. You can specify the name of this JAR file, the output directory, and the class package using Sqoop
command line options. For more information, see Sqoop documentation.
Getting information about the sqoop command
Use the help option of the sqoop import command to get online help on Sqoop command line options.
About this task
Use the help option of the sqoop import command to get online help on Sqoop command line options. For
example, on the Mac:
$ cd install_location/bin
$ ./dse sqoop import --help
100
DSE Search with Solr
DSE Search with Solr
DSE Search topics.
Getting started with Solr in DataStax Enterprise
DataStax Enterprise supports Open Source Solr (OSS) tools and APIs, simplifying migration from Solr to
DataStax Enterprise.
About this task
DataStax Enterprise supports Open Source Solr (OSS) tools and APIs, simplifying migration from Solr to
DataStax Enterprise. DataStax Enterprise is built on top of Solr 4.3.
DataStax Enterprise 4.0 turns off virtual nodes (vnodes) by default. DataStax does not recommend
turning on vnodes for Hadoop or Solr nodes, but you can use vnodes for any Cassandra-only cluster, or
a Cassandra-only data center in a mixed Hadoop/Solr/Cassandra deployment. If you have enabled virtual
nodes in Solr nodes, see Disabling virtual nodes. You can skip this step to run the Solr getting started
tutorial.
Introduction to Solr
The Apache Lucene project, Solr features robust full-text search, hit highlighting, and rich document
(PDF, Microsoft Word, and so on) handling. Solr also provides more advanced features like aggregation,
grouping, and geo-spatial search. Today, Solr powers the search and navigation features of many of the
world's largest Internet sites. With the latest version of Solr, near real-time indexing can be performed.
Solr integration into DataStax Enterprise
The unique combination of Cassandra, DSE Search with Solr, and DSE Analytics with Hadoop bridges the
gap between online transaction processing (OLTP) and online analytical processing (OLAP). DSE Search
in Cassandra offers a way to aggregate and look at data in many different ways in real-time using the
Solr HTTP API or the Cassandra Query Language (CQL). Cassandra speed can compensate for typical
MapReduce performance problems. By integrating Solr into the DataStax Enterprise big data platform,
DataStax extends Solr’s capabilities.
DSE Search is easily scalable. You add search capacity to your cluster in the same way as you add
Hadoop or Cassandra capacity to your cluster. You can have a hybrid cluster of nodes, provided the Solr
nodes are in a separate data center, some running Cassandra, some running search, and some running
101
DSE Search with Solr
Hadoop. If you have a hybrid cluster of nodes, follow instructions on isolating workloads. If you don't need
Cassandra or Hadoop, migrate to DSE strictly for Solr and create an exclusively Solr cluster. You can use
one data center and run Solr on all nodes.
Storage of indexes and data
The solr implementation in DataStax Enterprise does not support JBOD mode. Indexes are stored on the
local disk inside Lucene, actual data is stored in Cassandra.
Sources of information about OSS
Covering all the features of OSS is beyond the scope of DataStax Enterprise documentation. Because
DSE Search/Solr supports all Solr tools and APIs, refer to Solr documentation for information about topics,
such as how to construct Solr query strings to retrieve indexed data.
•
•
•
•
•
•
•
Apache Solr documentation
Solr Tutorial on the Solr site
Solr Tutorial on Apache Lucene site
Solr data import handler
Comma-Separated-Values (CSV) file importer
JSON importer
Solr cell project, which includes a tool for importing data from PDFs
For more information, see Solr 4.x Deep Dive by Jack Krupansky.
Benefits of using Solr in DataStax Enterprise
Solr offers real-time querying of files. Search indexes remain tightly in line with live data. There are
significant benefits of running your enterprise search functions through DataStax Enterprise instead of
OSS, including:
•
•
•
•
•
•
•
•
•
•
A fully fault-tolerant, no-single-point-of-failure search architecture
Linear performance scalability--add new search nodes online
Automatic indexing of data ingested into Cassandra
Automatic and transparent data replication
Isolation of all real-time, Hadoop, and search/Solr workloads to prevent competition between workloads
for either compute resources or data
The capability to read/write to any Solr node, which overcomes the Solr write bottleneck
Selective updates of one or more individual fields (a full re-index operation is still required)
Search indexes that can span multiple data centers (OSS cannot)
Solr/search support for CQL for querying Solr documents (Solr HTTP API may also be used)
Creation of Solr indexes from existing tables created with CQL
Data added to Cassandra is locally indexed in Solr and data added to Solr is locally indexed in Cassandra.
Supported and unsupported features
Supported and unsupported DSE Search and Solr features.
CQL-backed Solr cores require a new type mapping version 2. A CQL table must be created in Cassandra
before creating the Solr core. The schema corresponding to a CQL table using a compound primary key
requires a special syntax.
Unsupported Cassandra features
DSE Search does not support:
•
•
•
102
Cassandra 2.0.6 static columns
Cassandra compound primary keys for COMPACT STORAGE tables
Cassandra counter columns
DSE Search with Solr
•
Cassandra super columns
Unsupported Solr features
•
•
•
•
•
•
Solr schema fields that are both dynamic and multivalued for CQL-backed Solr cores (only)
The deprecated replaceFields request parameters on document updates for CQL-backed Solr cores.
Use the suggested procedure for inserting/updating data.
Block joins based on the Lucene BlockJoinQuery in Solr indexes and CQL tables
Schemaless mode
Schema updates through the REST API
org.apache.solr.spelling.IndexBasedSpellChecker and org.apache.solr.spelling.FileBasedSpellChecker
(org.apache.solr.spelling.DirectSolrSpellChecker is supported for spellchecking)
The commitWithin parameter
The SolrCloud CloudSolrServer feature of SolrJ for endpoint discovery and round-robin load balancing
•
•
Other unsupported features
DSE Search/Solr does not support JBOD mode.
Defining key Solr terms
Solr terms include several names for an index of documents and configuration on a single node.
In a distributed environment, such as DataStax Enterprise and Cassandra, the data is spread over multiple
nodes. In Solr, there are several names for an index of documents and configuration on a single node:
•
•
•
A Solr core
A collection
One shard of a collection
Each document in a Solr core/collection is considered unique and contains a set of fields that adhere to a
user-defined schema. The schema lists the field types and how they should be indexed. DSE Search maps
Solr cores/collections to Cassandra tables. Each table has a separate Solr core/collection on a particular
node. Solr documents are mapped to Cassandra rows, and document fields to columns. The shard is
analogous to a partition of the table. The Cassandra keyspace is a prefix for the name of the Solr core/
collection and has no counterpart in Solr.
This table shows the relationship between Cassandra and Solr concepts:
Cassandra
Solr--single node environment
Solr--distributed environment
Table
Solr core or collection
Collection
Row
Document
Document
Partition key
Unique key
Unique key
Column
Field
Field
Node
N/A
Node
Partition
N/A
Shard
Keyspace
N/A
N/A
With Cassandra replication, a Cassandra node or Solr core contains more than one partition (shard) of
table (collection) data. Unless the replication factor equals the number of cluster nodes, the Cassandra
node or Solr core contains only a portion of the data of the table or collection.
103
DSE Search with Solr
Installing Solr nodes
Installing and starting Solr nodes.
About this task
To install a Solr node, use the same installation procedure as you use to install any other type of node. To
use real-time (Cassandra), analytics (Hadoop), or search (Solr) nodes in the same cluster, segregate the
different nodes into separate data centers. Using the default DSESimpleSnitch automatically puts all the
Solr nodes in the same data center. Use OpsCenter Enterprise to rebalance the cluster when you add a
node to the cluster.
Starting and stopping a Solr node
The way you start up a Solr node depends on the type of installation, tarball or packaged.
Tarball installation:
From the install directory, use this command to start the Solr node:
$ bin/dse cassandra -s
The Solr node starts up.
From the install directory, use this command to stop the Solr node
$ bin/dse cassandra-stop
Packaged installation:
Procedure
1. Enable Solr mode by setting this option in /etc/default/dse:
SOLR_ENABLED=1
2. Start the dse service using this command:
$ sudo service dse start
The Solr node starts up.
You stop a Solr node using this command:
$ sudo service dse stop
Solr getting started tutorial
Steps for setting up Cassandra and Solr for the tutorial.
About this task
Setting up Cassandra and Solr for this tutorial involves the same basic steps as setting up a typical DSE
Search/Solr application:
•
•
•
Create a Cassandra table.
Import data.
Create a search index.
These steps for setting up Cassandra and Solr are explained in detail in this tutorial. After completing the
setup, you use DSE Search/Solr to perform simple queries, sort the query results, and construct facet
queries.
In this tutorial, you use some sample data from a health-related census.
104
DSE Search with Solr
Setup
This setup assumes you started DataStax Enterprise in DSE Search/Solr mode and downloaded the
sample data and tutorial files. The tutorial files consist of a Solr schema that corresponds to the CQL table
definition, which uses a compound primary key. The partitioning key is the id column and the clustering
key is the age column. Also included in the schema are several copy fields and a multivalued field that are
used for the faceted search.
Procedure
1. Download the sample data and tutorial files.
2. Unzip the files you downloaded in the DataStax Enterprise installation home directory.
A solr_CQL3tutorial directory is created that contains the following files.
•
copy_nhanes.cql
•
The COPY command you use to import data
create_nhanes.cql
•
The Cassandra CQL table definition
nhanes52.csv
•
The CSV (comma separated value) data
schema.xml
•
The Solr schema
solrconfig.xml
The Solr configuration file
3. Take a look at these files using your favorite editor.
Create a Cassandra table
Create a Cassandra table as part of the basic tutorial.
Procedure
1. Start cqlsh, and create a keyspace. Use the keyspace.
cqlsh> CREATE KEYSPACE nhanes_ks WITH REPLICATION =
{'class':'NetworkTopologyStrategy', 'Solr':1};
cqlsh> USE nhanes_ks;
2. Copy the CQL table definition from the downloaded create_nhanes.cql file, and paste in on the
cqlsh command line.
This action creates the nhanes table in the nhanes_ks keyspace.
Import data
After you create a Cassandra table, import data to set up DSE Search for the tutorial.
Procedure
1. Copy the cqlsh COPY command from the downloaded copy_nhanes.cql file.
2. Paste the COPY command on the cqlsh command line, change the FROM clause to match the path to /
solr_CQL3tutorial/nhanes52.csv in your environment, and then run the command.
This action imports the data from the CSV file into the nhanes table in Cassandra.
Create a search index
Create a search index for the tutorial.
105
DSE Search with Solr
Procedure
On the command line in the solr_CQL3tutorial directory, upload the solrconfig.xml and
schema.xml to Solr, and create the Solr core named after the Cassandra table and keyspace,
nhanes_ks.nhanes.
$ cd install_location/solr_CQL3tutorial
$ curl http://localhost:8983/solr/resource/nhanes_ks.nhanes/solrconfig.xml
--data-binary @solrconfig.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl http://localhost:8983/solr/resource/nhanes_ks.nhanes/schema.xml -data-binary @schema.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl "http://localhost:8983/solr/admin/cores?
action=CREATE&name=nhanes_ks.nhanes"
You can now begin searching.
Exploring the Solr Admin
After creating the Solr core, you can check that the Solr index is working by using the browser-based Solr
Admin.
About this task
After creating the Solr core, you can check that the Solr index is working by using the browser-based Solr
Admin:
http://localhost:8983/solr/
Procedure
To explore the Solr Admin:
1. Click Core Admin. Unless you loaded other Solr cores, the path to the default Solr core,
nhanes_ks.nhanes, appears.
At the top of the Solr Admin console, the Reload, Reindex, and Full Reindex buttons perform
functions that correspond to RELOAD command options. If you modify the schema.xml or
solrconfig.xml, you use these controls to re-index the data.
2. Check that the numDocs value is 20,050. The number of Solr documents corresponds to the number of
rows in the CSV data and nhanes table you created in Cassandra.
3. In Core Selector, select the name of the Solr core, nhanes_ks.nhanes.
Selecting the name of the Solr core brings up additional items, such as Query, in the vertical navigation
bar.
106
DSE Search with Solr
You can learn more about the Solr Admin from the Overview of the Solr Admin UI.
Running a simple search
Using the Solr Admin query form.
About this task
To search the database, experienced users run Solr HTTP API queries in a browser or on the command
line using the curl utility. You can also use the Solr Admin query form. Using the query form has some
advantages for those new to Solr. The form contains text entry boxes for constructing a query and can
provide query debugging information.
Procedure
To get started searching the nhanes database:
1. In the Solr Admin, click Query.
A query form appears.
107
DSE Search with Solr
Notice that the form has a number of query defaults set up, including the select URL in RequestHandler and *:* in the main query parameter entry box--q.
2. Scroll down the form and click Execute Query.
The defaults select all the fields in all the documents, starting with row 0 and ending at row 10. The wt
parameter specifies the output format, xml by default. The output looks something like this:
108
DSE Search with Solr
Running a faceted search
A faceted search drills down to filter search results that are based on a category of data.
Procedure
Run a complex query and include facet parameters in the request.
1. In the Solr Admin query form, specify a family size of 9 in the main query parameter text entry box--q:
family_size:9
2. In sort, specify sorting by age in ascending order, youngest to oldest:
age asc
3. In fl (filter list), specify returning only age and family size in results:
age family_size
Results from the main query will include only data about families of 9.
4. Click the facet option.
Text entry boxes for entering facet parameter values appear.
5. In facet.field, type this value:
109
DSE Search with Solr
age
The number of people in each age group will appear toward the end of the query results.
6. Click Execute Query.
The numfound value shows that 186 families having nine members were found. The query results
include only results from the fields in the filter list, age and family_size.
7. Scroll to the end of the query form to see the facet results.
110
DSE Search with Solr
The facet results show 11 people of age 17, 10 of age 34, and so on.
You can learn more about faceting from Solr documentation.
Solr HTTP API tutorial
Use the Solr HTTP API to run search queries.
About this task
For serious searching, use the Solr HTTP API. The Solr Admin query form is limited, but useful for learning
about Solr, and can even help you get started using the Solr HTTP API. The form shows the queries in Solr
HTTP format at the top of the form. After looking at a few URLs, you can try constructing queries in Solr
HTTP format.
Procedure
To get started using the Solr HTTP API:
1. Scroll to the top of the form, and click the greyed out URL.
111
DSE Search with Solr
A page of output, independent of the query form, appears that you can use to examine and change the
URL. The URL looks like this:
http://localhost:8983/solr/nhanes_ks.nhanes/select?
q=family_size%3A9&sort=age+asc&fl=age+family_size
&wt=xml&indent=true&facet=true&facet.field=age
2. In the URL in the address bar, make these changes:
FROM:
q=family_size%3A9
&fl=age+family_size
TO:
q=age:[20+TO+40]
&fl=age+family_size+num_smokers
The modifed URL looks like this:
http://localhost:8983/solr/nhanes_ks.nhanes/select?
q=age:[20+TO+40]&sort=age+asc&fl=age+family_size+num_smokers
&wt=xml&indent=true&facet=true&facet.field=age
In the Solr Admin query form, you can use spaces in the range [20 TO 40], but in the URL, you need to
use URL encoding for spaces and special characters. For example, use + or %20 instead of a space,
[20+TO+40].
3. Use the modified URL to execute the query. Move to the end of the URL, and press ENTER.
The number of hits increases from 186 to 7759. Results show the number of smokers and family size of
families whose members are 20-40 years old. Facets show how many people fell into the various age
groups.
. . .
</doc>
</result>
<lst name="facet_counts">
<lst name="facet_queries"/>
<lst name="facet_fields">
<lst name="age">
<int name="23">423</int>
<int name="24">407</int>
<int name="31">403</int>
<int name="30">388</int>
112
DSE Search with Solr
<int name="40">382</int>
<int name="28">381</int>
<int name="27">378</int>
<int name="21">377</int>
<int name="33">377</int>
<int name="22">369</int>
<int name="29">367</int>
<int name="20">365</int>
<int name="32">363</int>
<int name="34">361</int>
<int name="36">361</int>
<int name="25">358</int>
<int name="26">358</int>
<int name="35">358</int>
<int name="38">353</int>
<int name="37">339</int>
<int name="39">291</int>
<int name="17">0</int>
. . .
4. Experiment with different Solr HTTP API URLs by reading documentation on the internet and trying
different queries using this sample database.
This tutorial introduced you to DSE Search/Solr basic setup and searching. Next, delve into DataStax
Enterprise documentation and the recommended Solr documentation.
Configuring Solr
Configure Solr Type mapping.
About this task
Configure Solr Type mapping and virtual nodes before starting to use DSE Search/Solr. Solr types are
mapped to Cassandra validators shown in the Solr type mapping table. Configure legacy mapping of Solr
types if you created indexes in DataStax Enterprise 3.0.x or earlier.
DataStax Enterprise 4.0 turns off virtual nodes (vnodes) by default. DataStax does not recommend
turning on vnodes for Hadoop or Solr nodes, but you can use vnodes for any Cassandra-only cluster, or
a Cassandra-only data center in a mixed Hadoop/Solr/Cassandra deployment. If you have enabled virtual
nodes on Solr nodes, disable virtual nodes.
Changing maxBooleanClauses
A change to the maxBooleanClauses parameter in the solrconfig.xml requires restarting nodes to make the
change effective. Merely reloading the Solr cores does not suffice for this parameter.
Configuring the update log
The Solr update log is not supported because Cassandra provides the functionality. If you still want to
configure the update log, or you are upgrading a core that is configured to contain the update log, add the
force="true" attribute to the configuration element as follows, upload the new Solr configuration, and reload
the core:
<updateLog force="true">...</updateLog>
Mapping of Solr types
Table of Solr types.
This table shows the current DataStax Enterprise mapping of Solr types to Cassandra validators.
Solr Type
Cassandra Validator
Description
BCDIntField
Int32Type
Binary-coded decimal (BCD) integer
113
DSE Search with Solr
Solr Type
Cassandra Validator
Description
BCDLongField
LongType
BCD long integer
BCDStrField
UTF8Type
BCD string
BinaryField
BytesType
Binary data
BoolField
BooleanType
True (1, t, or T) or False (not 1, t, or T)
ByteField
Int32Type
Contains an 8-bit number value
DateField
DateType
Point in time with millisecond precision
DoubleField
DoubleType
Double (64-bit IEEE floating point)
EnumType
Int32Type
A closed set having a pre-determined sort order
ExternalFileField
UTF8Type
Values from disk file
FloatField
FloatType
32-bit IEEE floating point
IntField
Int32Type
32-bit signed integer
LongField
LongType
Long integer (64-bit signed integer)
RandomSortField
UTF8Type
Dynamic field in random order
ShortField
Int32Type
Short integer
SortableDoubleField
DoubleType
Numerically sorted doubles
SortableFloatField
FloatType
Numerically sorted floating point
SortableIntField
Int32Type
Numerically sorted integer
SortableLongField
LongType
Numerically sorted long integer
StrField
UTF8Type
String (UTF-8 encoded string or Unicode)
TextField
UTF8Type
Text, usually multiple words or tokens
TrieDateField
DateType
Date field for Lucene TrieRange processing
TrieDoubleField
DoubleType
Double field for Lucene TrieRange processing
TrieField
see description
Same as any Trie field type
TrieFloatField
FloatType
Floating point field for Lucene TrieRange processing
TrieIntField
Int32Type
Int field for Lucene TrieRange processing
TrieLongField
LongType
Long field for Lucene TrieRange processing
UUIDField
UUIDType
Universally Unique Identifier (UUID)
LatLonType
UTF8Type
Latitude/Longitude 2-D point, latitude first
PointType
UTF8Type
Arbitrary n-dimensional point for spatial search
GeoHashField
UTF8Type
Geohash lat/lon pair represented as a string
For efficiency in operations such as range queries, using Trie types is recommended. Keep the following
information in mind about these types:
•
UUIDField
DataStax Enterprise 4.0.2 and later supports the Cassandra TimeUUID type. A value of this type is a
Type 1 UUID that includes the time of its generation. Values are sorted, conflict-free timestamps. For
114
DSE Search with Solr
•
example, use this type to identify a column, such as a blog entry, by its timestamp and allow multiple
clients to write to the same partition key simultaneously.
BCD
•
A relatively inefficient encoding that offers the benefits of quick decimal calculations and quick
conversion to a string.
SortableDoubleField/DoubleType
•
If you use the plain types (DoubleField, IntField, and so on) sorting will be lexicographical instead of
numeric.
TrieField
Used with a type attribute and value: integer, long, float, double, date.
Mapping of CQL collections
DSE Search maps collections as follows:
•
•
Collection list and set: multi-valued field
Collection maps: dynamic field
The name of the dynamic field minus the wildcard is the map name. For example, a map column name
dyna* is mapped to dyna. Inner keys are mapped to the full field name.
Legacy mapping of Solr types
Mapping for DataStax Enterprise 3.0.x and earlier clusters.
DataStax Enterprise 3.0.x and earlier use the legacy type mapping by default.
Solr Type
Cassandra Validator
TextField
UTF8Type
StrField
UTF8Type
LongField
LongType
IntField
Int32Type
FloatField
FloatType
DoubleField
DoubleType
DateField
UTF8Type
ByteField
BytesType
BinaryField
BytesType
BoolField
UTF8Type
UUIDField
UUIDType
TrieDateField
UTF8Type
TrieDoubleField
UTF8Type
TrieField
UTF8Type
TrieFloatField
UTF8Type
TrieIntField
UTF8Type
TrieLongField
UTF8Type
All Others
UTF8Type
115
DSE Search with Solr
If you use legacy type mappings, the solr schema needs to define the unique key as a string.
Configuring the Solr type mapping version
The Solr type mapping version defines how Solr types are mapped to Cassandra Thrift or Cassandra
types.
The Solr type mapping version defines how Solr types are mapped to Cassandra Thrift or Cassandra
types, and plays an important role in upgrades too. While DataStax Enterprise 3.0 used a simplified
type mapping, also known as legacy mapping, or version 0, DataStax Enterprise 3.1 introduced a more
accurate type mapping with version 1.
During and after upgrades from 3.0.x to 3.1.x or 3.2.x to 4.0, tables created with DataStax Enterprise 3.0.x
requires legacy mapping, while new tables created with DataStax Enterprise 3.1.x and 3.2.x can use type
mapping version 1.
DataStax Enterprise 3.2 introduced a new type mapping, known as version 2 type mapping, to model the
latest CQL 3 changes. Tables migrated from previous DataStax Enterprise installations can keep the old
mapping versions, while newly created non-CQL3 tables require type mapping version 1, and new tables,
including those using compact storage, created using CQL 3 require type mapping version 2. CQL 3 is the
default mode in Cassandra 2.x, which is based on CQL specification 3.1.0.
To change the type mapping, configure dseTypeMappingVersion in the solrconfig.xml:
<dseTypeMappingVersion>2</dseTypeMappingVersion>
Set the value to 1 or 0 to enable one of the other versions. Switching between versions is not
recommended after the Solr core has been created successfully: attempting to load a solrconfig.xml with a
different dseTypeMappingVersion configuration and reloading the Solr core will cause an error.
Changing Solr Types
Changing a Solr type is not recommended. However, for particular circumstances, such as converting Solr
types such as the Solr LongField to TrieLongField, you configure the dseTypeMappingVersion using the
force option.
Changing a Solr type is rarely if ever done and is not recommended; however, for particular circumstances,
such as converting Solr types such as the Solr LongField to TrieLongField, you configure the
dseTypeMappingVersion using the force option.
The Cassandra internal validation classes of the types you are converting to and from must be compatible.
Also, the actual types you are converting to and from must be valid types. For example, converting
a legacy Trie type to a new Trie type is invalid because corresponding Cassandra validators are
incompatible. The output of the CLI command, DESCRIBE keyspace_name, shows the validation classes
assigned to columns.
For example, the org.apache.cassandra.db.marshal.LongType column validation class is mapped
to solr.LongType. You can force this column to be of the TrieLongField type by using force="true" in the
solrconfig.xml, and then performing a Solr core reload with re-indexing.
<dseTypeMappingVersion force = "true">1</dseTypeMappingVersion>
Use this option only if you are an expert and have confirmed that the Cassandra internal validation classes
of the types involved in the conversion are compatible.
To use DSE Search/Solr data from an 3.0 release or earlier, you need to use the legacy type mapping.
Configuring search components
Basic configuration for the Search Handler.
The wikipedia demo solrconfig.xml configures the Search Handler as follows:
<requestHandler name="search" class="solr.SearchHandler" default="true">
DataStax recommends using this basic configuration for the Search Handler.
116
DSE Search with Solr
Configuring additional search components
To configure the search handler for managing additional search components, you generally need to add
the additional component to the array of last-components to preserve the default configured components.
Distributed search does not work properly if you fail to preserve the default configured components. Unless
otherwise specified in Solr documentation, declare the additional component as described in the following
example.
How to add a search component
This example shows the configuration of an additional search component for spellchecking and how to add
that component to the last-components array of the search handler. The additional component specifies
the Java spelling checking package JaSpell:
Component configuration
<searchComponent class="solr.SpellCheckComponent" name="suggest_jaspell">
<lst name="spellchecker">
<str name="name">suggest</str>
<str name="classname">org.apache.solr.spelling.suggest.Suggester</str>
<str
name="lookupImpl">org.apache.solr.spelling.suggest.jaspell.JaspellLookup</
str>
<str name="field">suggest</str>
<str name="storeDir">suggest</str>
<str name="buildOnCommit">true</str>
<float name="threshold">0.0</float>
</lst>
</searchComponent>
To add the spell check component to the last-components array:
Last-components declaration
<requestHandler class="org.apache.solr.handler.component.SearchHandler"
name="/suggest">
<lst name="defaults">
<str name="spellcheck">true</str>
<str name="spellcheck.dictionary">suggest</str>
<str name="spellcheck.collate">true</str>
<str name="spellcheck.extendedResults">true</str>
</lst>
<arr name="last-components">
<str>suggest_jaspell</str>
</arr>
</requestHandler>
Configuring multithreaded DocValuesFacets
Setting the query executor threads parameter in the solrconfig.xml file.
You can set the query executor threads parameter in the solrconfig.xml file to enable multithreading
for filter queries, normal queries, and doc values facets.
<queryExecutorThreads>4</queryExecutorThreads>
Configuring the schema
A description of the Solr schema at a high level.
This document describes the Solr schema at a high level. For details about all the options and Solr schema
settings, see the Solr wiki. A Solr schema defines the relationship between data in a table and a Solr core.
The schema identifies the columns to index in Solr and maps column names to Solr types.
DataStax Enterprise supports CQL tables using simple primary keys, compound primary keys, as shown in
the Solr query join example, and composite partition keys.
117
DSE Search with Solr
Compound primary and composite partition keys
The Solr tutorial presents a schema for a Cassandra table that uses a CQL compound primary key. A CQL
table must be created in Cassandra before creating the Solr core. The schema for such a table requires a
different syntax than the simple primary key.
•
•
•
•
List each compound primary key column that appears in the CQL table in the Solr schema as a field,
just like any other column.
Declare the unique key using the key columns enclosed in parentheses.
Order the keys in the uniqueKey element as the keys are ordered in the CQL table.
When using composite partition keys, do not include the extra set of parentheses in the Solr uniqueKey.
Use a single set of parentheses and list the fields in the same order as you define the fields in CQL:
Cassandra Partition Key
CQL Syntax
Solr uniqueKey Syntax
Simple CQL primary key
CREATE TABLE ( . . . <a>
<type> PRIMARY KEY, . . . );
<uniqueKey>(a)</uniqueKey>
Compound primary key
CREATE TABLE ( . . . PRIMARY
KEY ( a, b, c ) );
<uniqueKey>(a, b, c)</
uniqueKey>
Composite partition key
CREATE TABLE ( . . . PRIMARY
KEY ( ( a, b), c );
<uniqueKey>(a, b, c)</
uniqueKey>
DSE Search/Solr maps schema fields and the unique key specification to the Cassandra key components,
and generates a synthetic unique key for Solr. The schema used by the tutorial is a synthetic unique
key that corresponds to the compound primary key in the Cassandra table definition, as shown in these
excerpts from the tutorial table and schema.xml:
Table definition
CREATE TABLE nhanes (
"id" INT,
"num_smokers" INT,
"age" INT,
. . .
PRIMARY KEY ("id", "age")
);
Schema definition
<schema name="solr_quickstart" version="1.1">
<types>
. . .
<fields>
<field name="id" type="int" indexed="true" stored="true"/>
<field name="num_smokers" type="int" indexed="true" stored="true"/>
<field name="age" type="int" indexed="true" stored="true"/>
. . .
<uniqueKey>(id,age)</uniqueKey>
. . .
Defining the unique key
The schema must have a unique key and must not duplicate rows. The unique key is like a primary key
in SQL. The unique key maps to the Cassandra partition key, which DataStax Enterprise uses to route
documents to cluster nodes.
The last element in the following sample schema names the unique key id. In a DSE Search/Solr schema,
the value of the stored attribute of non-unique fields needs to be true; True causes the field to be stored in
Cassandra. Solr indexes the field if indexed=true. An indexed field is searchable, sortable, and facetable.
Tokenized fields cannot be used as primary keys.
118
DSE Search with Solr
If you use legacy type mappings, the Solr schema needs to define the unique key as a string.
Sample schema
The following sample schema from the example of using a CQL collection set uses a simple primary key.
The schema specifies a StrCollectionField for quotes, a collection set column in the CQL table. A tokenizer
determines the parsing of the example text. The set of fields specifies the data that Solr indexes and
stores. DSE Search/Solr indexes the id, quotes, name, and title fields.
<schema name="my_search_demo" version="1.1">
<types>
<fieldType class="solr.StrField" multiValued="true"
name="StrCollectionField"/>
<fieldType name="string" class="solr.StrField"/>
<fieldType name="text" class="solr.TextField"/>
<fieldType class="solr.TextField" name="textcollection"
multiValued="true">
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
</analyzer>
</fieldType>
</types>
<fields>
<field name="id" type="string" indexed="true" stored="true"/>
<field name="quotes" type="textcollection" indexed="true" stored="true"/
>
<field name="name" type="text" indexed="true" stored="true"/>
<field name="title" type="text" indexed="true" stored="true"/>
</fields>
<defaultSearchField>quotes</defaultSearchField>
<uniqueKey>id</uniqueKey>
</schema>
Internal structure of the _uniqueKey field
In the Solr schema, you enclose the unique keys in parentheses if the field is a compound primary key
or composite partition key column in Cassandra. During indexing, DataStax Enterprise recognizes and
indexes the parenthetical as a _uniqueKey field. The structure of the _uniqueKey field is a string. The
value is structured as a JSON array of string elements. Types, such as booleans, are enclosed in quotation
marks. The actual type of the field is unimportant. Only the uniqueness of the value is important.
The structure of the _uniqueKey field is flat. The Cassandra-Solr-_uniqueKey mapping is:
Key
Cassandra
Solr
uniqueKey
Compound primary key
(a, b)
(a, b)
["a", "b"]
Composite partition key
((a, b), c)
(a, b, c)
["a", "b", "c"]
The final mapping to the uniqueKey flattens the Cassandra composite partition key ((a, b), c) on the Solr
side.
Changing a schema
Changing the Solr schema makes reloading the Solr core necessary. Re-indexing can be disruptive. Users
can be affected by performance hits caused by re-indexing. Changing the schema is recommended only
when absolutely necessary. Also, changing the schema during scheduled down time is recommended.
119
DSE Search with Solr
Limitations
DSE Search/Solr cannot index a document that indexes only one field, which is also the unique key in the
schema and the primary key in the corresponding Cassandra table. DSE Search/Solr deletes any existing
data with that primary key and it does not return any results for such a query.
Configuring the Solr library path
Relative paths are not supported. The workaround is to place custom code or Solr contrib modules in the
directories described in this topic.
Contrary to the examples shown in the solrconfig.xml indicating that relative paths are supported,
DataStax Enterprise does not support the relative path values set for the <lib> property. DSE Search/Solr
fails to find files placed in directories defined by the <lib> property. The workaround is to place custom
code or Solr contrib modules in these directories:
•
•
Packaged installs: /usr/share/dse
Binary installs: install_location/resources/dse/lib
Configuring the Data Import Handler
You can import data into DSE Search/Solr from data sources, such as XML and RDBMS using a
configuration-driven method that differs from the method used by open source Solr (OSS) to import data.
About this task
You can import data into DSE Search/Solr from data sources, such as XML and RDBMS. You use a
configuration-driven method that differs from the method used by open source Solr (OSS) to import data.
Requirements for using the Data Import Handler in DSE Search/Solr are:
•
•
A JDBC driver, the JDBC connection URL format, and driver class name for accessing the data source
for the data to be imported
Credentials for accessing the data to be imported
Procedure
1. Put the driver in the following DSE Search/Solr location and add the path to the driver to your PATH
environment variable.
• Tarball installs: install_location/resources/dse/lib
• Packaged installs: /usr/share/dse/solr/
2. Create a file named dataimport.properties that contains the following settings, modified for your
environment. Comment, uncomment, or edit the self-descriptive settings. The URL params section
refers to a mandatory suffix for the Solr HTTP API dataimport command.
# to sync or not to sync
# 1 - active; anything else - inactive
syncEnabled=1
#
#
which cores to schedule
in a multi-core environment you can decide which cores you want
synchronized
# leave empty or comment it out if using single-core deployment
#syncCores=coreHr,coreEn
# solr server name or IP address
# [defaults to localhost if empty]
server=localhost
# solr server port
# [defaults to 80 if empty]
port=8983
120
DSE Search with Solr
# application name/context
# [defaults to current ServletContextListener's context (app) name]
webapp=solrTest_WEB
# URL params [mandatory]
# remainder of URL
params=/select?qt=/dataimport&command=delta-import&clean=false&commit=true
# schedule interval
# number of minutes between two runs
# [defaults to 30 if empty]
interval=10
3. Save the dataimport.properties file in the following location:
•
Tarball installs:
•
install_location/resources/solr/conf
Packaged installs:
/etc/dse/cassandra/
4. Create a Solr schema to represent the data in Solr. For example:
<?xml version="1.0" encoding="UTF-8" ?>
<schema name="my_imported_data" version="1.0">
<types>
<fieldType name="text" class="solr.TextField">
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
</analyzer>
</fieldType>
<fieldType name="float" class="solr.FloatField" multiValued="false"/>
<fieldType name="int" class="solr.IntField" multiValued="false"/>
</types>
<fields>
<field name="mytable_key" type="int" indexed="true" stored="true"/>
<field name="myfield" type="int" indexed="true" stored="true"/>
. . .
</fields>
<uniqueKey>mytable_key</uniqueKey>
</schema>
5. Create a file named data-config.xml that maps the data to be imported to the Cassandra table that
is created automatically. For example:
<dataConfig>
<propertyWriter dateFormat="yyyy-MM-dd HH:mm:ss" type=
"SimplePropertiesWriter" directory=
"<install_location>/resources/solr/conf/" filename=
"dataimport.properties" />
<dataSource driver="org.mysql.jdbc.Driver" url=
"jdbc:mysql://localhost/mydb" user=
"changeme" password="changeme" />
<document name="test">
<entity name="cf" query="select * from mytable">
<field column="mytable_key" name="mytable_key" />
<field column="myfield" name="myfield" />
. . .
</entity>
</document>
</dataConfig>
6. Create a directory in the DataStax Enterprise installation home directory. Save the data-config.xml
in the directory you created.
7. From the following location, copy the solrconfig.xml.
121
DSE Search with Solr
• Tarball installs: install_location/demos/wikipedia
• Packaged installs: /usr/share/dse-demos/wikipedia
8. Paste the solrconfig.xml to the directory you created in step 6.
9. Add a requestHandler element to the solrconfig.xml file that contains the location of dataconfig.xml and data source connection information. For example:
<requestHandler name="/dataimport"
class="org.apache.solr.handler.dataimport.DataImportHandler">
<lst name="defaults">
<str name="config">data-config.xml</str>
<lst name="datasource">
<str name="driver">com.mysql.jdbc.Driver</str>
<str name="url">jdbc:mysql://localhost/mydb</str>
<str name="user">changeme</str>
<str name="password">changeme</str>
</lst>
</lst>
</requestHandler>
10.Upload the schema.xml, solrconfig.xml, and data-config.xml, and create the Solr core. For
example:
$ curl http://localhost:8983/solr/resource/mydb.mytable/solrconfig.xml -data-binary @solrconfig.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl http://localhost:8983/solr/resource/mydb.mytable/schema.xml --databinary @schema.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl http://localhost:8983/solr/resource/mydb.mytable/schema.xml --databinary @data-config.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl "http://localhost:8983/solr/admin/cores?
action=CREATE&name=mydb.mytable"
11.Import the data from the data source using HTTP API syntax. For example:
http://localhost:8983/solr/mydb.mytable/dataimport?command=full-import
where mydb is the Cassandra keyspace and mytable is the Cassandra table.
Creating a Solr index
Requirements and steps for creating a Solr index.
About this task
A minimal Solr installation requires these files:
•
Schema.xml
•
Describes the fields to index in Solr and types associated with them. These fields map to Cassandra
columns. To route search requests to the appropriate nodes, the schema needs a unique key.
Solrconfig.xml
Holds configuration information for query handlers and Solr-specific caches.
After writing a schema.xml, you HTTP-post the solrconfig.xml and the schema.xml to a Solr node in the
DataStax Enterprise cluster. Next, you create a new Solr core (or reload an existing core) to create (or
recreate) an index on a table for searching Cassandra data.
When users post schema or configuration files simultaneously, schema disagreements can occur. This
causes Solr errors.
122
DSE Search with Solr
Note: Do not make schema changes on hot production systems.
Uploading the schema and configuration
Steps to create a Solr index by posting the solrconfig.xml and schema.xml.
About this task
This procedure describes how to create a Solr index by posting the solrconfig.xml and schema.xml and
creating the Solr core. You can follow similar steps using an example to create a Solr index and insert data
into Solr and Cassandra.
Procedure
1. Post the configuration file using the cURL utility:
curl http://localhost:8983/solr/resource/keyspace.table/solrconfig.xml
--data-binary @solrconfig.xml -H 'Content-type:text/xml; charset=utf-8'
2. Post the schema file:
curl http://localhost:8983/solr/resource/keyspace.table/schema.xml
--data-binary @schema.xml -H 'Content-type:text/xml; charset=utf-8'
Creating a Solr core
Requirements and command for creating a Solr core.
About this task
You cannot create a Solr core unless you first upload the schema and configuration files. If you are
creating a CQL-backed Solr core, the table must be pre-exist in Cassandra before creating the core.
Use the curl command to create a Solr core.
$ curl "http://localhost:8983/solr/admin/cores?
action=CREATE&name=keyspace.table"
Creating a Solr core on one node automatically creates the core on other Solr nodes, and DSE Search
stores the files on all the Cassandra nodes.
By default, the cassandra user has full permissions on all keyspaces. If you specify a non-default user to
create a Solr core, the specified user must have the necessary Cassandra permissions. When specifying a
user to create a Solr core, ensure that the user has either:
•
•
CREATE permission on all keyspaces
All permissions on all keyspaces (superuser)
Reloading a Solr core
Reload a Solr core instead of creating a new one when you modify the schema.xml or solrconfig.xml.
Reload a Solr core instead of creating a new one when you modify the schema.xml or
solrconfig.xml.
$ curl "http://localhost:8983/solr/admin/cores?
action=RELOAD&name=keyspace.table"
You can use options with the RELOAD command to re-index and keep, or delete, the Lucene index. When
you do not specify an option, the default is used.
When you make a change to the schema, the compatibility of the existing index and the new schema
is questionable. If the change to the schema made changes to a field's type, the index and schema
will certainly be incompatible. Changes to a field's type can actually occur in subtle ways, occasionally
without a change to the schema.xml file itself. For example, a change to other configuration files, such as
synonyms, can change the schema. If such an incompatibility exists, a full re-index, which includes deleting
123
DSE Search with Solr
all the old data, of the Solr data is required. In these cases, anything less than a full re-index renders the
schema changes ineffective. Typically, a change to the Solr schema requires a full re-indexing.
Use these RELOAD command options to specify the level of re-indexing that occurs:
•
distributed
True, the default, distributes an index to nodes in the cluster. False re-indexes the Solr data on one
node. The false setting is used in certain recovery and upgrade procedures.
•
$ curl -v "http://localhost:8983/solr/admin/cores?action=RELOAD&
name=keyspace.table&distributed=false"
reindex and deleteAll
Re-indexes data in place or re-indexes in full. The default for both options is false. Accepting the
defaults reloads the core and no re-indexing occurs.
Re-indexing in place
Setting reindex=true and deleteAll=false re-indexes data and keeps the existing lucene index.
During the uploading process, user searches yield inaccurate results. To perform an in-place re-index, use
this syntax:
curl "http://localhost:8983/solr/admin/cores?action=RELOAD
&name=keyspace.table&reindex=true&deleteAll=false"
Re-indexing in full
Setting reindex=true and deleteAll=true deletes the Lucene index and re-indexes the dataset.
User searches initially return no documents as the Solr cores reload and data is re-indexed.
Setting reindex=false and deleteAll=true does nothing and generates an exception.
Rebuilding an index using the UI
You can re-index manually using the Solr Admin UI.
You can re-index manually using the UI or command-line tools. In the Core Admin screen of the Solr
Admin UI, the Reload, Reindex and Full Reindex buttons perform functions that correspond to RELOAD
command options.
Checking indexing status
If you HTTP post the files to a pre-existing table, DSE Search starts indexing the data. If you HTTP post
the files to a non-existent column keyspace or table, DSE Search creates the keyspace and table, and then
starts indexing the data.
If you HTTP post the files to a pre-existing table, DSE Search starts indexing the data. If you HTTP post
the files to a non-existent column keyspace or table, DSE Search creates the keyspace and table, and then
starts indexing the data. For example, you can change the stopwords.txt file, repost the schema, and
the index updates.
To check the indexing status, open the Solr Admin and click Core Admin.
124
DSE Search with Solr
You can also check the logs to get the indexing status. For example, you can check information about the
plugin initializer:
INDEXING / REINDEXING INFO SolrSecondaryIndex plugin initializer. 2013-08-26 19:25:43,347
SolrSecondaryIndex.java (line 403) Reindexing 439171 keys for core wiki.solr
Or you can check the SecondaryIndexManager.java information:
INFO Thread-38 2013-08-26 19:31:28,498 SecondaryIndexManager.java (line 136)
Submitting index build of wiki.solr for data in SSTableReader(path='/mnt/
cassandra/data/wiki/solr/wiki-solr-ic-5-Data.db'), SSTableReader(path='/mnt/
cassandra/data/wiki/solr/wiki-solr-ic-6-Data.db')
FINISH INDEXING INFO Thread-38 2013-08-26 19:38:10,701 SecondaryIndexManager.java (line 156)
Index build of wiki.solr complete
Adding and viewing index resources
Search includes a REST API for viewing and adding resources associated with an index.
DSE Search includes a REST API for viewing and adding resources associated with an index. You can
look at the contents of the existing Solr resource by loading its URL in a web browser or using HTTP get.
Retrieving and viewing resources returns the last uploaded resource, even if the resource is not the one
currently in use. If you upload a new schema, and then before reloading, request the schema resource,
Solr returns the new schema even though the Solr core continues to use the old schema.
Use this URL to post a file to Solr:
http://host:port/solr/resource/keyspace.table/filename.ext
Generally, you can post any resource required by Solr to this URL. For example, stopwords.txt and
elevate.xml are optional, frequently-used Solr configuration files that you post using this URL.
Using DSE Search/Solr
A brief description and illustration of DSE Search.
125
DSE Search with Solr
About this task
When you update a table using CQL, the Solr document is updated. Re-indexing occurs automatically after
an update.
Writes are durable. A Solr API client writes data to Cassandra first, and then Cassandra updates
indexes. All writes to a replica node are recorded both in memory and in a commit log before they are
acknowledged as a success. If a crash or server failure occurs before the memory tables are flushed to
disk, the commit log is replayed on restart to recover any lost writes.
The commit log replaces the Solr updatelog, which is not supported in DSE Search/Solr. Consequently,
features that require the updateLog are not supported:
•
•
•
Atomic updates
Real-time get
Versioning and optimistic concurrency
If you still want to use the update log, configure the updateLog in the solrconfig.xml using the force="true"
attribute.
126
DSE Search with Solr
Inserting, indexing, and searching data
Basic opertations.
About this task
The following examples show you how to perform basic operations:
•
•
Using a CQL collection set
Inserting data into Solr, creating a Solr core, and querying the data:
•
•
•
•
•
Executing CQL statements on the command line or from a client. Use the syntax described in the
DataStax Apache Cassandra documentation and shown in the following example of inserting data
into a CQL collection set.
• Using the Solr HTTP API update command
• Using any Thrift API, such as Pycassa or Hector
Deleting Solr data
Using dynamic fields Using dynamic fields to insert data
Using copy fields
Using docValues and copy fields for faceting
Example: Using a CQL collection set
An example of creating a table containing a CQL collection set of famous quotations.
About this task
In this example, you create a table containing a CQL collection set of famous quotations. You insert data
into the table by copying/pasting INSERT commands from a file that you download. Download the INSERT
commands in the quotations.zip file now.
Next, you index the data in DSE Search/Solr, and finally, query the table using the Solr HTTP API.
Procedure
1. If you did not already create a directory named solr_CQL3tutorial that contains a schema.xml and
solrconfig.xml, do so now. Copy the schema.xml and solrconfig.xml from the demos/
wikipedia directory to solr_CQL3tutorial.
2. After starting DSE as a Solr node, open a shell window or tab, go to the bin directory on Linux for
example, and start CQL:
./cqlsh
3. Create a keyspace and a table consisting of a set collection column and other columns, and then, insert
some data for DSE Search to index.
CREATE KEYSPACE mykeyspace
WITH REPLICATION = {'class':'NetworkTopologyStrategy', 'Solr':1};
USE mykeyspace;
CREATE TABLE mysolr (
id text PRIMARY KEY,
name text,
title text,
quotes set <text>
);
4. Unzip the quotations.zip file that you downloaded, copy the insert commands, and paste them on
the cqlsh command line.
5. Change the schema.xml in the solr_CQL3 tutorial directory to contain the schema shown in the
Sample schema section.
127
DSE Search with Solr
6. Create a search index using the cURL utility. On the operating system command line in the
solr_CQL3tutorial directory, post the solrconfig.xml and the schema.xml, and create a Solr
core.
$curl http://localhost:8983/solr/resource/mykeyspace.mysolr/solrconfig.xml
--data-binary @solrconfig.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl http://localhost:8983/solr/resource/mykeyspace.mysolr/schema.xml -data-binary @schema.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl "http://localhost:8983/solr/admin/cores?
action=CREATE&name=mykeyspace.mysolr"
If you are recreating the mykeyspace.mysolr core, use the RELOAD instead of the CREATE command.
7. Search Cassandra using the Solr HTTP API to find titles like Succ*.
http://localhost:8983/solr/mykeyspace.mysolr/
select?q=title%3ASucc*&wt=json&indent=on&omitHeader=on
{
"response":{"numFound":2,"start":0,"docs":[
{
"id":"126",
"title":"Success",
"quotes":["If A is success in life, then A equals x plus y
plus z. Work is x; y is play; and z is keeping your mouth
shut."],
"name":"Albert Einstein"},
{
"id":"125",
"title":"Success",
"quotes":["Always bear in mind that your own resolution to
succeed is more important than any one thing.",
"Better to remain silent and be thought a fool than to speak
out and remove all doubt."],
"name":"Abraham Lincoln"}]
}}
Inserting/updating data using the Solr HTTP API
Steps for Inserting/updating a CQL-based core using the Solr HTTP API.
About this task
To update indexed data in Solr and in the Cassandra table, use a URL in the following format to update the
document:
curl http://host:port/solr/keyspace.table/update?
replacefields=false -H 'Content-type: application/json' -d
'json string'
Updates to a CQL-backed Solr core replace the entire row. The deprecated replacefields parameter for
inserting into, modifying, or deleting data from CQL Solr cores is not supported.
Procedure
Building on the collections example, insert data into the mykeyspace.mytable data and Solr index.
Use this curl command might look like this:
$ curl http://localhost:8983/solr/mykeyspace.mysolr/update?
replacefields=false -H 'Content-type: application/json' -d '[{"id":"130",
"body":"Life is a beach.", "name":"unknown", "title":"Life"}]'
128
DSE Search with Solr
The Solr convention is to use curl for issuing update commands instead of using a browser. You
do not have to post a commit command in the update command as you do in OSS, and doing so is
ineffective.
When you use CQL or CLI to update a field, DSE Search implicitly sets replacefields to false and
updates individual fields in the Solr document. The re-indexing of data occurs automatically.
Warning about using the optimize command
Do not include the optimize command in URLs to update Solr data. This warning appears in the system
log when you use the optimize:
WARN [http-8983-2] 2013-03-26 14:33:04,450
CassandraDirectUpdateHandler2.java (line 697)
Calling commit with optimize is not recommended.
The Lucene merge policy is very efficient. Using the optimize command is no longer necessary and
using the optimize command in a URL can cause nodes to fail.
Using dynamic fields
Using dynamic fields, you can index content in fields that are not explicitly defined by the schema.
About this task
Using dynamic fields, you can index content in fields that not explicitly defined by the schema. Using
dynamic fields, you can also process multiple Solr fields the same way by using a generic prefix or suffix to
reference the field. A common use case for dynamic fields is to catch fields that should not be indexed or
to implement a schema-less index. As previously mentioned, in CQL-backed Solr cores Solr schema fields
that are dynamic and multivalued are not supported.
To use a dynamic field:
•
Include a Solr dynamic field in the schema.xml.
Name the field using wildcard at the beginning or end of the field. For example, an asterisk prefix or
suffix in the field name in the schema designates a dynamic field.
•
•
• dyna_*
• *_s
In CQL, to define the map collection column, use the same base name (no asterisk) as you used for the
field in the schema.xml.
For example, use dyna_* in the schema.xml and dyna_ for the name of the CQL map colleciton.
Using CQL, insert data into the map using the base name as a prefix or suffix in the first component of
each map pair. The format of the map using a prefix is:
{ prefix_literal : literal, prefix_literal : literal, . . . }
DSE Search maps the Solr dynamic field to a Cassandra map collection column
Example: A multilingual music database
Procedure
This example demonstrates how to use dynamic fields to organize song data based on language into a
collection map.
1. Use the keyspace from the collection set example.
2. In cqlsh, create the following table:
CREATE TABLE hits (
song uuid,
lang_ map<text, text>,
129
DSE Search with Solr
PRIMARY KEY (song)
);
3. If you did not already create a directory named solr_CQL3tutorial that contains a schema.xml and
solrconfig.xml, do so now. You can use the schema.xml and solrconfig.xml from the demos/
wikipedia directory by copying these files to solr_CQL3tutorial.
4. Change the schema.xml in the solr_CQL3tutorial directory to contain this schema:
<schema name="topHits" version="1.1">
<types>
<fieldType name="uuid" class="solr.UUIDField" />
<fieldType name="text" class="solr.TextField">
</fieldType>
</types>
<fields>
<field name="song" type="uuid" indexed="true" stored="true"/>
<dynamicField name="lang_*" type="text" indexed="true" stored="true"/>
</fields>
<defaultSearchField>song</defaultSearchField>
<uniqueKey>song</uniqueKey>
</schema>
5. Create a search index using the cURL utility. On the operating system command line in the
solr_CQL3tutorial directory, post the configuration file, the schema file, and create a Solr core.
curl http://localhost:8983/solr/resource/mykeyspace.hits/solrconfig.xml -data-binary @solrconfig.xml -H 'Content-type:text/xml; charset=utf-8'
curl http://localhost:8983/solr/resource/mykeyspace.hits/schema.xml --databinary @schema.xml -H 'Content-type:text/xml; charset=utf-8'
curl "http://localhost:8983/solr/admin/cores?
action=CREATE&name=mykeyspace.hits"
6. Using CQL, insert the following data about Italian and Hawaiian songs into the hits table. Use the lang_
to prefix the first component of each map pair.
INSERT INTO hits (song, lang_) VALUES
( 62c36092-82a1-3a00-93d1-46196ee77204, { 'lang_i-title' : 'La Vita E La
Felicita', 'lang_i-artist' : 'Michele Bravi' });
INSERT INTO hits (song, lang_) VALUES ( 8a172618-b121-4136-bb10f665cfc469eb, { 'lang_h-title' : 'Blew it', 'lang_h-artist' : 'Maoli f/
Fiji' });
INSERT INTO hits (song, lang_) VALUES ( a3e64f8f-bd44-4f28b8d9-6938726e34d4, { 'lang_i-title' : 'Dimmi Che Non Passa Felicita',
'lang_i-artist' : 'Violetta' });
7. To find data about hit songs in Italy, query on either of the prefixed map literals, lang_i-title or lang_iartist.
http://localhost:8983/solr/mykeyspace.hits/select?q=lang_i-title
%3A*&wt=xml&indent=true
Results are:
<result name="response" numFound="2" start="0">
<doc>
<str name="song">62c36092-82a1-3a00-93d1-46196ee77204</str>
<str name="lang_i-artist">Michele Bravi</str>
<str name="lang_i-title">La Vita E La Felicita</str.</doc>
<doc>
<str name="song">a3e64f8f-bd44-4f28-b8d9-6938726e34d4</str>
<str name="lang_i-artist">Violetta</str>
<str name="lang_i-title">Dimmi Che Non Passa Felicita</str></doc>
</result>
130
DSE Search with Solr
Deleting Solr data
To delete a Cassandra table and its data, including the data indexed in Solr, from a Solr node, drop the
table using CQL.
About this task
To delete a Cassandra table and its data, including the data indexed in Solr, from a Solr node, drop the
table using CQL. The following example, which assumes you ran the example of using a collection set, lists
the Solr files on the file system, drops the table named mysolr that the demo created, and then verifies that
the files have been deleted from the file system:
Wait until you've finished working through all the examples in this document before actually deleting the
example data.
Procedure
1. List the Solr data files on the file system.
•
Packaged installs:
•
ls /usr/local/var/lib/dse5/data/solr.data/mykeyspace.mysolr/index/
Tarball installs:
ls /var/lib/cassandra/data/solr.data/mykeyspace.mysolr/index
The output looks something like this:
_33.fdt
_35_nrm.cfe
_38_Lucene40_0.tim
_33.fdx
_35_nrm.cfs
_38_Lucene40_0.tip
_33.fnm
_36.fdt
_38_nrm.cfe
. . .
2. Launch cqlsh and execute the CQL command to drop the table named solr.
DROP TABLE mykeyspace.mysolr;
3. Exit cqlsh and check that the files have been deleted on the file system. For example:
ls /var/lib/cassandra/data/solr.data/mykeyspace.mysolr/index
The output is:
ls: /var/lib/cassandra/data/solr.data/mykeyspace.mysolr/index: No such file
or directory
Using copy fields
Search supports the stored=false copy field directive in the schema.xml.
About this task
The way DSE Search/Solr handles copy fields depends on the value of the stored attribute.
If stored=false in the copyField directive:
•
•
Ingested data is copied by the copyField mechanism to the destination field for search, but data is not
stored in Cassandra.
When you add a new copyField directive to the schema.xml, pre-existing and newly ingested data is
re-indexed when copied as a result of the new directive.
If stored=true in the copyField directive:
•
•
Ingested data is copied by the copyField mechanism and data is stored in Cassandra.
When you add a new copyField directive to the schema.xml, pre-existing data is re-indexed as the
result of an old copyField directive, but not when copied as the result of a new copyField directive. To
be re-indexed, data must be re-ingested after you add a new copyField directive to the schema.
131
DSE Search with Solr
DataStax Enterprise 4.0.3 and later supports stored copy fields having different source and destination
data types.
Using a copy field and multivalued field
When you use copy fields to copy multiple values into a field, CQL comes in handy because you do
not need to format the data in json, for example, when you insert it. Using the Solr HTTP API update
command, the data must be formatted.
Use the CQL BATCH command to insert column values in a single CQL statement to prevent overwriting.
This process is consistent with Solr HTTP APIs, where all copied fields need to be present in the inserted
document. You need to use BATCH to insert the column values whether or not the values are stored in
Cassandra.
Using docValues and copy fields for faceting
Using docValues can improve performance of faceting, grouping, filtering, sorting, and other operations
described on the Solr Wiki.
For faceting to use docValues, the schema needs to specify multiValued="true" even if the field is a singlevalue facet field. The field needs to include docValues="true". You also need to use a field type that
supports being counted by Solr. The text type, which tokenizes values, cannot be used, but the string type
works fine. DataStax Enterprise supports all aspects of copy fields except:
•
•
The maxChars attribute is not supported.
Copying from/to the same dynamic field is not supported.
Example: copy fields and docValues
This example uses copy fields to copy various aliases, such as a twitter name and email alias, to a
multivalue field. You can then query the multivalue field using any alias as the term to get the other aliases
in the same row or rows as the term.
About this task
This example uses copy fields to copy various aliases, such as a twitter name and email alias, to a
multivalue field. You can then query the multivalue field using any alias as the term to get the other aliases
in the same row or rows as the term.
Procedure
1. If you did not already create a directory named solr_CQL3tutorial that contains a schema.xml
and solrconfig.xml, do so now. You can use the schema.xml and solrconfig.xml from the
demos/wikipedia directory by copying these files to solr_CQL3tutorial.
2. Using CQL, create a keyspace and a table to store user names, email addresses, and their skype,
twitter, and irc names. The all field will exist in the Solr index only, so you do not need an all column in
the table.
CREATE KEYSPACE user_info
WITH REPLICATION = { 'class' : 'SimpleStrategy', 'replication_factor' :
1 };
CREATE TABLE user_info.users (
id text PRIMARY KEY,
name text,
email text,
skype text,
irc text,
twitter text
) ;
3. Run a CQL BATCH command, as explained earlier, if the schema includes a multivalue field.
BEGIN BATCH
132
DSE Search with Solr
INSERT INTO user_info.users (id, name, email, skype, irc, twitter) VALUES
('user1', 'john smith', '[email protected]', 'johnsmith', 'smitty',
'@johnsmith')
INSERT INTO user_info.users (id, name, email, skype, irc, twitter) VALUES
('user2', 'elizabeth doe', '[email protected]', 'roadwarriorliz',
'elizdoe', '@edoe576')
INSERT INTO user_info.users (id, name, email, skype, irc, twitter) VALUES
('user3', 'dan graham', '[email protected]', 'danielgra', 'dgraham',
'@dannyboy')
INSERT INTO user_info.users (id, name, email, skype, irc, twitter) VALUES
('user4', 'john smith', '[email protected]', 'johnsmith', 'jsmith345',
'@johnrsmith')
INSERT INTO user_info.users (id, name, email, skype, irc, twitter) VALUES
('user5', 'john smith', '[email protected]', 'jdsmith', 'jdansmith',
'@smithjd999')
INSERT INTO user_info.users (id, name, email, skype, irc, twitter) VALUES
('user6', 'dan graham', '[email protected]', 'dangrah', 'dgraham',
'@graham222')
APPLY BATCH;
4. Use a schema that contains the multivalued field--all, copy fields for each alias plus the user id, and a
docValues option.
<schema name="my_search_demo" version="1.1">
<types>
<fieldType name="string" class="solr.StrField"/>
<fieldType name="text" class="solr.TextField">
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
</analyzer>
</fieldType>
</types>
<fields>
<field name="id" type="string" indexed="true" stored="true"/>
<field name="name" type="string" indexed="true" stored="true"/>
<field name="email" type="string" indexed="true" stored="true"/>
<field name="skype" type="string" indexed="true" stored="true"/>
<field name="irc" type="string" indexed="true" stored="true"/>
<field name="twitter" type="string" indexed="true" stored="true"/>
<field name="all" type="string" docValues="true" indexed="true"
stored="false" multiValued="true"/>
</fields>
<defaultSearchField>name</defaultSearchField>
<uniqueKey>id</uniqueKey>
<copyField source="id" dest="all"/>
<copyField source="email" dest="all"/>
<copyField source="skype" dest="all"/>
<copyField source="irc" dest="all"/>
<copyField source="twitter" dest="all"/>
</schema>
5. On the command line in the solr_CQL3tutorial directory, upload the schema.xml and
solrconfig.xml to Solr. Create the Solr core for the keyspace and table, user_info.users.
$ curl http://localhost:8983/solr/resource/user_info.users/solrconfig.xml
--data-binary @solrconfig.xml -H 'Content-type:text/xml; charset=utf-8'
$ curl http://localhost:8983/solr/resource/user_info.users/schema.xml
--data-binary @schema.xml -H 'Content-type:text/xml; charset=utf-8'
133
DSE Search with Solr
$ curl "http://localhost:8983/solr/admin/cores?
action=CREATE&name=user_info.users"
6. In a browser, search Solr to identify the user, alias, and id of users having an alias smitty.
http://localhost:8983/solr/user_info.users/select?q=all
%3Asmitty&wt=xml&indent=true
The output is:
<result name="response" numFound="1" start="0">
<doc>
<str name="id">user1</str>
<str name="twitter">@johnsmith</str>
<str name="email">[email protected]</str>
<str name="irc">smitty</str>
<str name="name">john smith</str>
<str name="skype">johnsmith</str>
</doc>
</result>
7. Run this query:
http://localhost:8983/solr/user_info.users/select/?
q=*:*&facet=true&facet.field=name&facet.mincount=1&indent=yes
At the bottom of the output, the facet results appear. Three instances of john smith, two instances of
dan graham, and one instance of elizabeth doe:
. . .
</result>
<lst name="facet_counts">
<lst name="facet_queries"/>
<lst name="facet_fields">
<lst name="name">
<int name="john smith">3</int>
<int name="dan graham">2</int>
<int name="elizabeth doe">1</int>
</lst>
</lst>
. . .
8. Now you can view the status of the field cache memory to see the RAM usage of docValues per Solr
field. Results look something like the example shown in Example 2.
9. In the Solr Admin, after selecting a Solr core from the drop-down menu, click Plugins / Stats. Expand
dseFieldCache and dseFilterCache to see information about the per-segment field cache and filter
cache.
Choose Watch Changes or Refresh Values to get updated information.
134
DSE Search with Solr
Changing the copy field value
Steps for changing the value of the copy field.
Procedure
1. Change the stored attribute value of a copyField directive from true to false.
a) Change the values of stored copyField directives to false.
b) Post the solrconfig.xml and the modified schema.xml.
c) Reload the Solr core, specifying an in-place re-index.
Previously stored copies of data are not automatically removed from Cassandra.
2. Changing the stored attribute value from false to true is not directly supported. The workaround is:
a) Remove the copyField directives that have stored=false.
b) Reload the solrconfig.xml and schema.xml. Use the reindex=true option.
c) Add back the copyField directives you removed to the schema.xml and set stored=true.
d) Post the solrconfig.xml and the modified schema.xml.
e) Reload the Solr core, specifying an in-place re-index.
f) Re-ingest the data.
Stored values are not automatically removed from Cassandra.
Viewing the Solr core status
Using the Solr API to view the status of the Solr core.
About this task
You can use the Solr API to view the status of the Solr core. For example, to view the status of the wiki.solr
core after running the wikipedia demo, use this URL:
135
DSE Search with Solr
http://localhost:8983/solr/#/~cores/wiki.solr
136
DSE Search with Solr
137
DSE Search with Solr
Status of all Solr cores
To view the status of all Solr cores use this URL:
http://localhost:8983/solr/admin/cores?action=STATUS
For example, the status of the wiki.solr core looks like this:
{
"defaultCoreName":"default.1371321667755813000",
"initFailures":{},
"status":{
"wiki.solr":{
"name":"wiki.solr",
"isDefaultCore":false,
"instanceDir":"solr/",
"dataDir":"/var/lib/cassandra/data/solr.data/wiki.solr/",
"config":"solrconfig.xml",
"schema":"schema.xml",
"startTime":"2013-06-16T21:05:54.894Z",
"uptime":7212565,
"index":{
"numDocs":3579,
"maxDoc":3579,
"deletedDocs":0,
"version":532,
"segmentCount":15,
"current":false,
"hasDeletions":false,
"directory":"org.apache.lucene.store.
NRTCachingDirectory:NRTCachingDirectory
(org.apache.lucene.store.NIOFSDirectory
@/private/var/lib/cassandra/data/solr.data/wiki.solr/index
lockFactory=
[email protected];
maxCacheMB=48.0 maxMergeSizeMB=4.0)",
"userData":{"commitTimeMSec":"1371416801053"},
"lastModified":"2013-06-16T21:06:41.053Z",
"sizeInBytes":8429726,
"size":"8.04 MB"},
"indexing":false}}}
Status of field cache memory
The Solr field cache caches values for all indexed documents, which if left unchecked, can result in out-ofmemory errors. For example, when performing faceted queries using multi-valued fields the multiValued
fields are multi-segmented (as opposed to single segmented single-valued fields), resulting in an inefficient
near real time (NRT) performance. You can use densely packed DocValue field types and per-segment
docsets. Facet queries will be per-segment, which improves real-time search performance problems.
To ensure that the jvm heap can accommodate the cache, monitor the status of the field cache and take
advantage of the Solr 4.3 option for storing the cache on disk or on the heap. To view the status of the
field cache memory usage, append &memory=true to the URL used to view the status of Solr cores. For
example, to view the field cache memory usage of the DSE Search quick start example after running a few
facet queries, use this URL:
http://localhost:8983/solr/admin/cores?action=STATUS&memory=true
Example 1
For example, the URL for viewing the field cache memory usage in json format and the output is:
http://localhost:8983/solr/admin/cores?
action=STATUS&wt=json&indent=on&omitHeader=on
138
DSE Search with Solr
&memory=true
. . .
"memory":{
"unInvertedFields":{
"totalSize":0,
"totalReadableSize":"0 bytes"},
"multiSegment":{
"multiSegment":"StandardDirectoryReader(segments_3:532:nrt _6p(4.3):
C3193 _71(4.3):C161 _6i(4.3):C15 _6n(4.3):C21 _6e(4.3):C16 _6k(4.3):
C19 _6t(4.3):C17 _6g(4.3):C10 _77(4.3):C12 _6v(4.3):C9 _7c(4.3):
C66 _72(4.3):C14 _6x(4.3):C7 _6y(4.3):C7 _6w(4.3):C12)",
"fieldCache":{
"entriesCount":0},
"totalSize":0,
"totalReadableSize":"0 bytes"},
"segments":{
"_6p":{
"segment":"_6p",
"docValues":{
. . .
"fieldCache":{
"entriesCount":0},
"totalSize":51600,
"totalReadableSize":"50.4 KB"}},
"totalSize":619200,
"totalReadableSize":"604.7 KB"}},
"totalMemSize":619200,
"totalReadableMemSize":"604.7 KB"}}
Example 2
After running a few sort by query functions, the output looks something like this:
. . .
"fieldCache":{
"entriesCount":1,
"id":{
"cacheType":"org.apache.lucene.index.SortedDocValues",
"size":260984,
"readableSize":"254.9 KB"}},
"totalSize":260984,
"totalReadableSize":"254.9 KB"},
"segments":{
. . .
"fieldCache":{
"entriesCount":2,
"age":{
"cacheType":"int",
"size":3832,
"readableSize":"3.7 KB"},
"id":{
"cacheType":"int",
"size":3832,
"readableSize":"3.7 KB"}},
"totalSize":59232,
"totalReadableSize":"57.8 KB"}},
"totalSize":524648,
"totalReadableSize":"512.4 KB"}},
139
DSE Search with Solr
"totalMemSize":524648,
"totalReadableMemSize":"512.4 KB"}}
Using the field cache
In Lucene-Solr 4.5 and later, docValues are mostly disk-based to avoid the requirement for large heap
allocations in Solr. If you use the field cache in sort, stats, and other queries, make those fields docValues.
Querying Solr data
A brief description about query Solr data.
About this task
DSE Search hooks into the Cassandra Command Line Interface (CLI), Cassandra Query Language
(CQL) library, the cqlsh tool, existing Solr APIs, and Thrift APIs. Avoid querying nodes that are indexing.
For responding to queries, DSE Search ranks the nodes that are not performing Solr indexing higher
than indexing ones. If only indexing nodes can satisfy the query, the query will not fail but instead return
potentially partial results.
Using SolrJ and other Solr clients
About using SolrJ and other Solr clients.
Solr clients work with DataStax Enterprise. If you have an existing Solr application, using it with DataStax
Enterprise is straight-forward. Create a schema, then import your data and query using your existing
Solr tools. The Wikipedia demo is built and queried using Solrj. The query is done using pure Ajax. No
Cassandra API is used for the demo.
You can also use any Thrift API, such as Pycassa or Hector, to access DSE-Search. Pycassa supports
Cassandra indexes. You can use indexes in Pycassa just as you use the solr_query expression in DSE
Search.
DataStax has extended SolrJ to protect internal Solr communication and HTTP access using SSL. You can
also use SolrJ to change the consistency level of a DSE-Search node.
Shard selection
A brief description about how DSE search uses a shuffling technique and other strategies to improve
performance.
DataStax Enterprise uses a shuffling technique to balance the load, and also attempts to minimize the
number of shards queried as well as the amount of data transferred from non-local nodes.
Using the ShardRouter Mbean
Use the com.datastax.bdp:type=ShardRouter Mbean to retrieve information and update the list of
endpoints.
DataStax Enterprise exposes the com.datastax.bdp:type=ShardRouter Mbean, providing the following
operations:
•
•
•
•
140
getShardSelectionStrategy(String core) retrieves the name of the shard selection strategy used for the
given Solr core.
getEndpoints(String core) retrieves the list of endpoints that can be queried for the given Solr core.
getEndpointLoad(String core) retrieves the list of endpoints with related query load for the given Solr
core; the load is computed as a 1-minute, 5-minutes and 15-minutes exponentially weighted moving
average, based on the number of queries received by the given node.
refreshEndpoints() manually refreshes the list of endpoints to be used for querying Solr cores.
DSE Search with Solr
Using the Solr HTTP API
Using the Solr HTTP API to query data indexed in DSE Search/Solr.
The Solr HTTP API is preferred over CQL solr_query statements for querying Cassandra for correctness
and performance reasons; the solr_query is suitable only for simple, brief, and occasional testing and
for simple administrative tasks, not for production usage. You can use the Solr HTTP API to query data
indexed in DSE Search/Solr just as you would search for data indexed in OSS.
Solr HTTP API example
Assuming you performed the example of using a collection set, to find the titles in the mykeyspace.mysolr
table that begin with the letters Succ in XML, use this URL:
http://localhost:8983/solr/mykeyspace.mysolr/select?q=%20title
%3ASucc*&fl=title
The response is:
<response>
<lst name="responseHeader">
<int name="status">0</int>
<int name="QTime">2</int>
<lst name="params">
<str name="fl">title</str>
<str name="q">title:Succ*</str>
</lst>
</lst>
<result name="response" numFound="2" start="0">
<doc>
<str name="title">Success</str>
</doc>
<doc>
<str name="title">Success</str>
</doc>
</result>
</response>
Delete by query
After you issue a delete by query, documents start getting deleted immediately and deletions continue until
all documents are removed. For example you can delete the data that you inserted using this command on
the operating system command line:
$ curl http://localhost:8983/solr/mykeyspace.mysolr/update --data
'<delete><query>*:*</query></delete>' -H
'Content-type:text/xml; charset=utf-8'
Delete by id
Delete by id deletes the document with a specified id and is more efficient than delete by query.
Delete by id deletes the document with a specified id and is more efficient than delete by query. The id
is the value of the uniqueKey field declared in the schema. The id can be a synthetic id that represents a
Cassandra compound primary key, such as the one used in the Solr tutorial. To delete by id, the following
example builds on the example in running a simple search. After clicking Execute Query, a list of results
appears. Each result includes a _uniqueKey in JSON format. The uniqueKey is the first line of each result
and looks like this:
<str name="_uniqueKey">["47336","29"]</str>
In this example, ["47336", "29"] are the values of the id, age compound primary key. The following delete
by id query shows the HTTP API command you use to remove that particular record from the Solr index:
141
DSE Search with Solr
$ curl http://localhost:8983/solr/nhanes_ks.nhanes/update --data
'<delete><id>["47336","29"]</id></delete>' -H 'Content-type:text/xml;
charset=utf-8'
After deleting the record, run a simple search on the Solr tutorial data again. The Solr Admin shows that
the number of documents has been reduced by one: 20049. Query the Cassandra table using cqlsh:
cqlsh:nhanes_ks> SELECT * FROM nhanes WHERE id=47336;
The cqlsh output also confirms that the data was removed. Null values appear instead of the data.
Joining cores
You can join Solr documents, including those having different Solr cores under certain conditions.
DataStax Enterprise 4.0.2 and later supports the OS Solr query time join through a custom implementation.
You can join Solr documents, including those having different Solr cores under these conditions:
•
•
•
•
Solr cores need to have the same keyspace and same Cassandra partition key.
Both Cassandra tables that support the Solr cores to be joined have to be either Thrift- or CQLcompatible. You cannot have one that is Thift-compatible and one that is CQL-compatible.
The type of the unique key (Cassandra key validator of the partition key) are the same.
The order of table partition keys and schema unique keys are the same.
Re-indexing CQL-backed cores
The custom query time join adds a field to the index for joining on composite partition keys. This addition
necessitates re-indexing of pre-existing CQL-backed Solr cores. Thrift-backed cores do not need reindexing.
Simplified syntax
This simplified syntax is recommended for joining Solr cores:
q={!join fromIndex=test.from}field:value
The custom implementation eliminates the need to use to/from parameters required by OS Solr. Based on
the key structure, DataStax Enterprise can determine what the parameters are. For backward compatibility
with applications, the verbose, legacy syntax is also supported.
Example of using a query time join
This example creates two tables, songs and lyrics. The tables use the same partition key. The songs table
uses a simple primary key, the UUID of a song. The primary key of the songs table is its partition key. The
lyrics table uses a compound primary: id and song, both of type UUID. After joining cores, you construct a
single query to retrieve information about songs having lyrics that include "love".
You can copy CQL commands, Solr HTTP requests, and the query from the downloaded commands.txt
file.
1. Download and unzip the file containing the Solr schemas, Solr configuration files, and commands for
this example.
This action creates /songs and /lyrics directories, schemas, and Solr configuration files for indexing
data in the the songs and lyrics tables.
2. Start cqlsh, and then create and use a keyspace named internet.
You can copy/paste from the downloaded commands.txt file.
3. Create two tables, song and lyrics, that share the internet keyspace and use the same partition key.
cqlsh> CREATE TABLE songs (song uuid PRIMARY KEY, title text, artist
text);cqlsh> CREATE TABLE lyrics (song uuid, id uuid, lyrics text, PRIMARY
KEY (song, id));
142
DSE Search with Solr
Both tables share the song partition key, a uuid. The second table also contains the id clustering
column.
4. Insert the data from the downloaded file into the songs table.
5. Insert data into the lyrics table.
The lyrics of songs by Big Data and John Cedrick mention love.
6. Navigate to the songs directory that you created in step 1, and take a look at the Solr schema.xml.
Navigate to the lyrics directory and take a look at the schema. Notice that the order of the unique key in
the schema and the partition key of the lyrics table are the same: (song, id). Using (id, song) does not
work.
<schema name="songs_schema" version="1.1">
<types>
<fieldType name="uuid" class="solr.UUIDField" />
<fieldType name="text" class="solr.TextField">
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
</analyzer>
</fieldType>
</types>
<fields>
<field name="song" type="uuid" indexed="true" stored="true"/>
<field name="title" type="text" indexed="true" stored="true"/>
<field name="artist" type="text" indexed="true" stored="true"/>
</fields>
<defaultSearchField>artist</defaultSearchField>
<uniqueKey>song</uniqueKey>
</schema>
<schema name="lyrics_schema" version="1.1">
<types>
<fieldType name="uuid" class="solr.UUIDField" />
<fieldType name="text" class="solr.TextField" >
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
</analyzer>
</fieldType>
</types>
<fields>
<field name="song" type="uuid" indexed="true" stored="true"/>
<field name="id" type="uuid" indexed="true" stored="true"/>
<field name="words" type="text" indexed="true" stored="true"/>
</fields>
<defaultSearchField>words</defaultSearchField>
<uniqueKey>(song, id)</uniqueKey>
</schema>
7. In the songs directory, post the solrconfig.xml and schema.xml for the internet.songs core, and create
the Solr core for internet.songs.
8. In the lyrics directory, post the solrconfig.xml and schema.xml for the internet.lyrics core, and create the
Solr core for internet.lyrics.
9. Search for songs that have lyrics about love.
http://localhost:8983/solr/internet.songs/select/?q={!join
+fromIndex=internet.lyrics}words:love&indent=true&wt=json
The output includes two songs having the word "love" in the lyrics, one by Big Data and the other by
John Cedrick:
"response":{"numFound":2,"start":0,"docs":[
{
"song":"a3e64f8f-bd44-4f28-b8d9-6938726e34d4",
"title":"Dangerous",
143
DSE Search with Solr
"artist":"Big Data"},
{
"song":"8a172618-b121-4136-bb10-f665cfc469eb",
"title":"Internet Love Song",
"artist":"John Cedrick"}]
}}
Recursive join support
You can nest a join query to use the result of one join as an input for another join, and another, recursively.
All joined data must reside on the same partition. To embed one query in the Solr query string of another,
use the magic field name _query_.
Use the following syntax to construct a query that recursively joins cores.
F1:V1 AND _query_:"{!join fromIndex=keyspace.table}(F2:V2 AND _query_:\"{!join
fromIndex=keyspace.table}(F3:V3)\")"
Where the top level from query includes a nested join query. The nested join in this example is:
_query_:\"{!join fromIndex=keyspace.table}(F3:V3)\"
Like an SQL SELECT IN ... (SELECT IN ...) query, Solr executes the nested join queries first, enabling
multiple nested join queries if required.
A Solr join query is not a relational join where the values from the nested join queries are returned in the
results.
Example of a recursive join query
This example builds on the solr query time join example. Embed in the query to join songs and lyrics
having words:"love" a second query to join award-winning videos using AND _query_:"award:true".
You can copy CQL commands, Solr HTTP requests, and the query from the downloaded commands.txt
file.
1. In cqlsh, create a videos table that shares the internet keyspace and uses the same partition key as the
songs and lyrics tables.
cqlsh> CREATE TABLE videos (song uuid, award boolean, title text, PRIMARY
KEY (song));
2.
3.
4.
5.
All three tables use the song partition key, a uuid.
Insert the data from the downloaded file into the videos table. The video data sets the award field to true
for the videos featuring songs by Big Data and Brad Paisley.
Navigate to the videos directory that was created when you unzipped the downloaded file.
In the videos directory, post the solrconfig.xml and schema.xml, and create the Solr core for
internet.videos.
Use a nested join query to recursively join the songs and lyrics documents with the videos document,
and to select the song that mentions love and also won a video award.
http://localhost:8983/solr/internet.songs/select/?q=
{!join+fromIndex=internet.lyrics}words:love AND _query_:
{!join+fromIndex=internet.videos}award:true&indent=true&wt=json
Output is:
"response":{"numFound":1,"start":0,"docs":[
{
"song":"a3e64f8f-bd44-4f28-b8d9-6938726e34d4",
"title":"Dangerous",
"artist":"Big Data"}]
}}
144
DSE Search with Solr
Support for the legacy join query
DataStax Enterprise supports using the legacy syntax that includes to/from fields in the query. The
requirements for using the legacy syntax are:
•
•
Tables do not use composite partition keys.
The query includes the force=true local parser parameter, as shown in this example that joins mytable1
and mytable2 in mykeyspace.
Legacy syntax example
curl 'http://localhost:8983/solr/mykeyspace.mytable1/select/?q=\{!join+from=id
+to=id+fromIndex=mykeyspace.mytable2+force=true\}'
Tracing Solr HTTP requests
To troubleshoot queries, trace Solr HTTP requests.
For debugging and troubleshooting, you can trace Solr HTTP requests in one of the following ways:
•
•
Enable probabilistic tracing.
Pass an explicit cassandra.trace=true request parameter in the HTTP query.
After running the example of using a join query, you can trace the join query by adding the cassandra.trace
parameter to the HTTP request:
http://localhost:8983/solr/internet.songs/select/?
q={!join+from=song+to=id+fromIndex=internet.lyrics
+force=true}words:love&indent=true&wt=json&cassandra.trace=true
The Solr response includes a cassandra.trace.session value, the unique session id of the tracing session
for the request:
{
"cassandra.trace.session":"3e503490-bdb9-11e3-860f-73ded3cb6170",
"responseHeader":{
"status":0,
"QTime":1,
"params":{
"indent":"true",
"q":"{!join from=song to=id fromIndex=internet.lyrics
force=true}words:love",
"wt":"json",
"cassandra.trace":"true"}},
"response":{"numFound":2,"start":0,"docs":[
{
"id":"8a172618-b121-4136-bb10-f665cfc469eb",
"title":"Internet Love Song",
"artist":"John Cedrick"},
{
"id":"a3e64f8f-bd44-4f28-b8d9-6938726e34d4",
"title":"Dangerous",
"artist":"Big Data"}]
}}
To see the information from the trace, query the system_traces.events, using the session id to filter the
output.
cqlsh> select * from system_traces.events where session_id = 3e503490bdb9-11e3-860f-73ded3cb6170;
session_id
| activity
| source_elapsed
------------+--------------------------------------------------------------------------+----------------
145
DSE Search with Solr
3e503490... |
Parsing SELECT * from "internet"."songs" WHERE "id" =
8a172618...|
2607
3e503490... |
Preparing
statement |
3943
3e503490... |
Executing single-partition
query on songs |
4246
3e503490... |
Acquiring sstable
references |
4261
3e503490... |
Merging memtable
tombstones |
4305
3e503490... |
Key cache hit for
sstable 1 |
4388
3e503490... |
Seeking to partition indexed section in
data file |
4399
3e503490... | Skipped 0/1 non-slice-intersecting sstables, included 0 due to
tombstones |
4873
3e503490... |
Merging data from memtables and
1 sstables |
4954
3e503490... |
Read 1 live and 0
tombstoned cells |
5162
3e503490... |
Parsing SELECT * from "internet"."songs" WHERE "id" =
a3e64f8f... |
6160
3e503490... |
Preparing
statement |
7424
. . .
For example purposes, the event_id, node IP address, and thread id have been deleted from this output to
fit on the page.
In the case of distributed queries over several nodes, Cassandra uses the same tracing session id on
all nodes, which makes it possible to correlate Cassandra operations on all the nodes taking part in the
distributed query.
Limiting columns indexed and returned by a query
Changing the default column limit in Thrift tables.
When using dynamic fields, the default column limit controls the maximum number of indexed columns
overall, not just dynamic field columns. The column limit also controls the maximum number of columns
returned during queries. This column limit prevents out of memory errors caused by using too many
dynamic fields. If dynamic fields are not used, the column limit has no effect.
To change the default column limit, which is 1024, configure the dseColumnLimit element in the
solrconfig.xml file. You can override the default configuration using the column.limit parameter in a query to
specify a different value, for example 2048.
http://localhost:8983/solr/keyspace.table/select?q=
title%3Amytitle*&fl=title&column.limit=2048
Querying multiple tables
Using the Solr API to query multiple tables simultaneously having the same schema.
To map multiple Cassandra tables to a single Solr core, use the Solr HTTP API. Specify multiple tables
using the shards parameter. For example:
http://host:port/solr/keyspace1.cf1/select?q=*:*&shards=
host:port/solr/keyspace1.cf1,host:port/solr/keyspace2.cf2
Using the Solr API, you can query multiple tables simultaneously if they have same schema.
Querying using autocomplete/spellcheck
Query specifying the autocomplete/spellcheck behavior using the shards.qt= parameter.
146
DSE Search with Solr
By default, the solrconfig.xml does not include configuration for the Solr suggestor. After configuring
the search component in the solrconfig.xml for /suggest, you can issue a query specifying the
autocomplete/spellcheck behavior using the shards.qt= parameter. For example, to test the suggestor:
curl "http://localhost:8983/solr/mykeyspace.mysolr/select?shards.qt=/
suggest&qt=/suggest&q=testin"
Using CQL
DSE Search supports production-grade implementation of CQL Solr queries. You can develop CQL-centric
applications supporting full-text search without having to work with Solr-specific APIs.
About this task
You can use a solr_query expression in a SELECT statement to retrieve Solr data from Cassandra. CQL
tables, including compound primary keys and collections, are supported. CQL tables must already exist
before creating the Solr core. Legacy Thrift/CQL2 tables are automatically created from the Solr schema if
no table exists when creating the Solr core.
Synopsis
SELECT select expression
FROM table
[WHERE solr_query = 'search expression'] [LIMIT n]
<search expression> syntax is a Solr query string that conforms to the Lucene syntax and Solr query
syntax. You enclose the Solr query string in single quotation marks. For example, after running the
Wikipedia demo you can use these Solr query strings:
Type of Query
Example
Description
Field search
'title:natio* AND Kenya'You can use multiple fields defined in the
schema: 'title:natio* AND body:CarlosAragonés'
Wildcard search
'Ken?a'
Use ? or * for single or multi-character searches.
Fuzzy search
'Kenya~'
Use with caution, many hits can occur.
Phrase search
'"American football player"'
Searches for the phrase enclosed in double quotation
marks.
Proximity search
'"football Bolivia"~10' Searches for football and Bolivia within 10 words of each
other.
Range searches
'title:[football TO
soccer}'
Term boosting
'"football"^4 "soccer"' By default, the boost factor is 1. Must be a positive number.
Boolean operator
'+Macedonian
football'
AND, +, OR, NOT and - can be used.
Grouping
'(football OR
soccer) AND Carlos
Aragonés'
Use parentheses to group clauses.
Field grouping
'title:(+football
+"Bolivia")'
Use parentheses to group multiple clauses into one field.
Supports both inclusive and exclusive bounds using square
brackets and curly braces, respectively.
Tips for using solr_query
If you use CQL to query DSE Search and do not get a response, the likely cause is incorrect solr_query
syntax. In response to incorrect syntax, DSE Search simply returns no results and logs an error message
in the Cassandra log. Check the Cassandra log, which is located in /var/log/cassandra by default.
147
DSE Search with Solr
A SELECT expression reads one or more records from a Cassandra table and returns a result-set of rows.
Each row consists of a partition key and a collection of columns corresponding to the query. Unlike the
projection in a SQL SELECT, there is no guarantee that the results will contain all of the columns specified.
An error does not occur if you request non-existent columns. Check the Cassandra log for error messages.
LimitationIf you index a legacy table, which by definition uses the COMPACT STORAGE directive, in Solr
you cannot use the key alias, for example id, in a SELECT statement, for example SELECT id FROM . . .
Querying such a table returns an error because the table schema is updated from Thrift and the key alias is
lost. A SELECT key FROM . . . returns the correct result.
CQL Example
To query the Wikipedia demo search results:
Procedure
1. Connect to the cqlsh. On the Mac, for example:
cd install_location/bin
./cqlsh
2. Use the wiki keyspace and include the solr_query expression in a CQL select statement to find the titles
in the table named solr that begin with the letters natio:
USE wiki;
SELECT title FROM solr WHERE solr_query='title:natio*';
The output, sorted in lexical order, appears:
title
-------------------------------------------------------------------------Kenya national under-20 football team
Bolivia national football team 2002
Israel men's national inline hockey team
List of French born footballers who have played for other national teams
Bolivia national football team 1999
Bolivia national football team 2001
Bolivia national football team 2000
Using the ExtendedDisMax query parser
The ExtendedDisMax Query Parser (eDisMax) includes more features than than the traditional Solr query
parser, such as multi-field query (Disjunction Max Query), relevancy calculations, full query syntax, and
aliasing.
About this task
The traditional Solr query parser (defType=lucene) is the default query parser and intended for
compatibility with traditional Solr queries. The ExtendedDisMax Query Parser (eDisMax) includes more
features than than the traditional Solr query parser, such as multi-field query (Disjunction Max Query),
relevancy calculations, full query syntax, and aliasing. eDisMax is essentially a combination of the
traditional Solr query parser and the dismax query parser, plus a number of other functional and usability
enhancements. It is the most powerful query parser for Solr that is offered out of the box. For more
information, see Solr 4.x Deep Dive by Jack Krupansky.
eDisMax supports phrase query features, such as phrase fields and phrase slop. You can use full query
syntax to search multiple fields transparently, eliminating the need for inefficient copyField directives.
eDisMax example
To query the mykeyspace.mysolr table from collection set example using eDisMax, specify the edismax
deftype, and query the default field for either of two words: yearning or kills.:
148
DSE Search with Solr
http://localhost:8983/solr/mykeyspace.mysolr/
select?&defType=edismax&q=yearning or kills
&wt=json&indent=on&omitHeader=on
Output in json format is:
{
"response":{"numFound":2,"start":0,"docs":[
{
"id":"123",
"title":"Life",
"quotes":["Life is a foreign language; all men mispronounce it.",
"There are three ingredients in the good life: learning, earning
and yearning."],
"name":"Christopher Morley"
},
{
"id":"124",
"title":"Life",
"quotes":["In matters of self-control as we shall see again and
again, speed kills.
But a little friction really can save lives.", "We Have Met the
Enemy:
Self-Control in an Age of Excess."],
"name":"Daniel Akst"}]
}
}
Document level boosting
To add document-level boosting on CQL tables, add a column named _docBoost of type float to the table.
Fields belonging to that document will be boosted at indexing time.
Capacity planning
Use a discovery process to develop a plan to ensure sufficient memory resources.
About this task
Using DSE Search/Solr is memory-intensive. This discovery process is intended to help you, the DSE
Search/Solr administrator, develop a plan for having sufficient memory resources to meet the needs of
your users.
Overview
First, you estimate how large your Solr index will grow by indexing a number of documents on a single
node, executing typical user queries, and then examining the field cache memory usage for heap
allocation. Repeat this process using a greater number of documents until you get a feel for the size of the
index for the maximum number of documents that a single node can handle. You can then determine how
many servers to deploy for a cluster and the optimal heap size. The index should be stored on SSDs or
should fit in the system IO cache.
Although capacity planning requires a serious effort by operations personnel, the results justify the time
spent when planning achieves these results:
•
•
Optimal heap size per node
Good estimate about the number of nodes needed for your application.
The replication factor can be increased for more queries per second.
Note: The Pre-flight tool can detect and fix many invalid or suboptimal configuration settings.
149
DSE Search with Solr
Before you begin
You need to have the following hardware and data:
A node with:
•
•
GB of RAM, size to be determined during capacity planning
SSD or spinning disk
Input data:
•
•
•
N documents indexed on a single test node
A complete set of sample queries to be executed
The total number of documents the system should support
Capacity planning process
Procedure
1. Create schema.xml and solrconfig.xml files.
2. Start a node.
3. Add N docs.
4.
5.
6.
7.
8.
9.
Run a range of queries that simulate those of users in a production environment.
View the status of field cache memory to discover the memory usage.
View the size of the index (on disk) included in the status information about the Solr core.
Based on the server's system IO cache available, set a maximum index size per-server.
Based on the memory usage, set a maximum heap size required per-server.
Calculate the maximum number of documents per node based on #6 and #7.
When the system is approaching the maximum docs per-node, add more nodes.
Mixing workloads in a cluster
About using real-time (Cassandra), Hadoop, or search (Solr) nodes in the same cluster.
About this task
A common question is how to use real-time (Cassandra), analytics (Hadoop), or search (Solr) nodes in the
same cluster. Within the same data center, attempting to run Solr on some nodes and real-time queries or
analytics on other nodes does not work. The answer is to organize the nodes running different workloads
into virtual data centers.
Creating a virtual data center
Virtual data centers are a convenient way to organize work loads within clusters. When you create a
keyspace using CQL, you can set up virtual data centers, independent of what physical data center the
individual nodes are in. You assign analytics nodes to one data center, search nodes to another, and
Cassandra real-time nodes to yet another data center. The separate, virtual data centers for different types
of nodes segregate workloads running Solr from those running Cassandra real-time or Hadoop analytics
applications. Segregating workloads ensures that only one type of workload is active per data center.
Workload segregation
The batch needs of Hadoop and the interactive needs of Solr are incompatible from a performance
perspective, so these workloads need to be segregated. Cassandra real-time applications and DSE
Search/Solr applications are also incompatible, but for a different reason--dramatically distinct access
patterns:
150
DSE Search with Solr
•
A Cassandra real-time application needs very rapid access to Cassandra data.
•
The real-time application accesses data directly by key, large sequential blocks, or sequential slices.
A DSE Search/Solr application needs a broadcast or scatter model to perform full-index searching.
Virtually every Solr search needs to hit a large percentage of the nodes in the virtual data center
(depending on the RF setting) to access data in the entire Cassandra table. The data from a small
number of rows are returned at a time.
In this diagram, nodes in data centers 1 and 2 (DC 1 and DC 2) can run a mix of:
•
•
Real-time queries (Cassandra and no other services)
Analytics (Cassandra and Hadoop)
Data centers 3 and 4 (DC 3 and DC 4) are dedicated to search.
In separate data centers, some DSE nodes handle search while others handle MapReduce, or just act as
real-time Cassandra nodes. Cassandra ingests the data, Solr indexes the data, and you run MapReduce
against that data, all in one cluster without having to do any manual extract, transform, and load (ETL)
operations. Cassandra handles the replication and isolation of resources.
The Solr nodes run HTTP and hold the indexes for the Cassandra table data. If a Solr node goes down,
the commit log replays the Cassandra inserts, which correspond to Solr inserts, and the node is restored
automatically.
The diagram shows the different types of node, even Analytics and Cassandra nodes, in separate data
centers. Separating Analytics and Cassandra nodes is not mandatory as it is between Solr and the other
node types. Separating all workloads is typically, but not always, recommended by DataStax. Occasionally,
there is a use case for keeping Analytics and Cassandra in the same data center. You do not have to have
an one or more additional replication factors when nodes are in the same data center.
To deploy a mixed workload cluster, see Multiple data center deployment.
Restrictions
•
•
•
For production use or for use with mixed workloads, do not create the keyspace using SimpleStrategy.
SimpleStrategy will not work in mixed workload clusters running Solr. NetworkTopologyStrategy is
highly recommended for most deployments because it is much easier to expand to multiple data
centers when required by future expansion.
Do not run Solr and Hadoop on the same node in either production or development environments.
From Hadoop and Cassandra real-time clusters in multiple data centers, do not attempt to insert data
to be indexed by Solr using CQL or Thrift. Run the CQL or Thrift inserts on a Solr node in its own data
center.
151
DSE Search with Solr
Replicating data across data centers
You set up replication for Solr nodes exactly as you do for other nodes in a Cassandra cluster, by creating
a keyspace. You can change the replication of a keyspace after creating it.
Common operations
Topics for using DSE Search.
About this task
You can run Solr on one or more nodes. DataStax does not support running Solr and Hadoop on the same
node, although it's possible to do so in a development environment. In production environments, separate
workloads by running real-time (Cassandra), analytics (Hadoop), or DSE Search (Solr) nodes on separate
nodes and in separate data centers.
Handling inconsistencies in query results
Consider session stickiness, subrange node repair, and follow best practices for soft commit points on
different replica nodes.
DSE Search implements an efficient, highly available distributed search algorithm on top of Cassandra,
which tries to select the minimum number of replica nodes required to cover all token ranges, and also
avoid hot spots. Consequently, due to the eventually consistent nature of Cassandra, some replica nodes
might not have received or might not have indexed the latest updates yet. This situation might cause DSE
Search to return inconsistent results (different numFound counts) between queries due to different replica
node selections. This behavior is intrinsic to how highly available distributed systems work, as described
in the ACM article, "Eventually Consistent" by Werner Vogels. Most of the time, eventual consistency is
not an issue, yet DSE Search implements session stickiness to guarantee that consecutive queries will hit
the same set of nodes on an healthy, stable cluster, hence providing monotonic results. Session stickiness
works by adding a session seed to request parameters as follows:
shard.shuffling.strategy=SEED
shard.shuffling.seed=<session id>
In the event of unstable clusters with missed updates due to failures or network partitions, consistent
results can be achieved by repairing nodes using the subrange repair method.
Finally, another minor source of inconsistencies is caused by different soft commit points on different
replica nodes: A given item might be indexed and committed on a given node, but not yet on its replica.
This situation is primarily a function of the load on each node, hence DataStax recommends the following
practices:
•
•
•
Evenly balancing read/write load between nodes
Properly tuning soft commit time and async indexing concurrency
Configuring back pressure in the dse.yaml file
Adding, decommissioning, repairing a node
About adding, removing, or repairing a Solr node.
To increase the number of nodes in a Solr cluster, you can add a DSE node to the cluster. If you want to
increase capacity of your search, add the node, then optionally, rebalance the cluster. To add a Solr node,
use the same method you use to add a Cassandra node. Using the default DSESimpleSnitch automatically
puts all the Solr nodes in the same data center. Use OpsCenter Enterprise to rebalance the cluster.
Decommissioning and repairing a node
You can decommission and repair a Solr node in the same manner as you would a Cassandra node. The
efficient and recommended way to repair a node, or cluster, is to use the subrange repair method.
152
DSE Search with Solr
Shuffling shards to balance the load
Strategies for load balancing.
To balance the load in a distributed environment, you can choose from several strategies for shuffling the
shards to suit your needs. The shard shuffling strategy specifies how one node is selected over others for
reading the Solr data. There are several methods for selecting the strategy. All methods involve setting the
shard.shuffling.strategy parameter to one of the following values:
Possible values of shard.shuffling.strategy
•
host
•
Shards are selected based on the host that received the query
query
•
Shards are selected based on the query string.
host_query
•
Shards are selected by host x query.
random
•
Different random set of shards are selected with each request (default).
SEED
Selects the same shard from one query to another.
Methods for selecting shard shuffling strategy
•
Append shard.shuffling.strategy = <strategy> to the HTTP API query. For example:
http://localhost:8983/solr/wiki.solr/select?
q=title:natio*&shard.shuffling.strategy=host
•
Issuing this query determines the shard shuffling strategy for this query only.
Create a dse-search.properties file and POST it to Solr.
For example:
1. Create the dse-search.properties file having the following contents:
shard.shuffling.strategy=query
2. Post the command to DSE Search/Solr. For example:
curl -v --data-binary @dse-search.properties
http://localhost:8983/solr/resource/wiki.solr/dse-search.properties
•
Posting the command determines the shard shuffling strategy for all queries to the given Solr core.
The strategy is propagated to all nodes and saved in Solr core metadata.
Set the following parameters to use the SEED strategy:
1. Pass the shard.shuffling.strategy=SEED as a request parameter.
2. Specify a request parameter, such as an IP address or any string, using the
shard.shuffling.seed parameter. When you reuse the same seed value between queries on a
stable cluster, the same shard strategy will be in effect.
Every time you pass the same string, the same list of shards is queried, regardless of the target
node you actually query; if you change the string, a different list of shards are queried.
3. Verify that the strategy was maintained by passing the shards.info=true request parameter.
For example:
curl "http://localhost:8983/solr/demo.solr/select/?
q=text:search&shards.info=true&shard.shuffling.strategy=SEED&shard.shuffling.seed=19
Shuffling does not always result in the node selection you might expect. For example, using a replication
factor of 3 with six nodes, the best and only solution is a two-shard solution--half of the data read from the
originator node and half from another node. A three-shard solution would be inefficient.
153
DSE Search with Solr
Managing the location of Solr data
Controlling where the Solr data files are saved on the server.
Solr has its own set of data files. Like Cassandra data files, you can control where the Solr data files are
saved on the server. By default, the data is saved in Cassandra data directory/solr.data. You
can change the location from the Cassandra data directory to another directory, from the command line.
For example, on Linux:
cd install_directory
bin/dse cassandra -s -Ddse.solr.data.dir=/opt
In this example, the solr dta is saved in the /opt directory.
Solr log messages
DSE Search logs Solr errors, warnings, debug, trace, and info messages in the Cassandra system log.
DSE Search logs Solr errors, warnings, debug, trace, and info messages in the Cassandra system log:
/var/log/cassandra/system.log
Changing the Solr logging level
Assuming you configured and are using the Apache log4j utility, you can control the granularity of Solr
log messages, and other log messages, in the Cassandra system.log file by configuring the log4jserver.properties file. The log4j-server.properties file is located in:
•
•
Packaged installs: : /etc/dse/cassandra
Binary Installs: /resources/cassandra/conf/
To set log levels, configure the log4j.rootLogger value, specifying one of these values:
•
•
•
•
•
•
•
•
All - turn on all logging
OFF - no logging
FATAL - severe errors causing premature termination
ERROR - other runtime errors or unexpected conditions
WARN - use of deprecated APIs, poor use of API, near errors, and other undesirable or unexpected
runtime situations
DEBUG - detailed information on the flow through the system
TRACE - more detailed than DEBUG
INFO - highlight the progress of the application at a coarse-grained level
For example, open the log4j-server.properties file and change the log level by configuring the
log4j.rootLogger value:
# output messages into a rolling log file as well as stdout
log4j.rootLogger=INFO,stdout
Accessing the validation Log
DSE Search stores validation errors that arise from non-indexable data sent from non-Solr nodes in this
log:
/var/log/cassandra/solrvalidation.log
For example, if a Cassandra node that is not running Solr puts a string in a date field, an exception is
logged for that column when the data is replicated to the Solr node.
Changing the Solr connector port
Changing the Solr port from the default port.
To change the Solr port from the default, 8983, change the http.port setting in the
catalina.properties file installed with DSE in install_location/resources/tomcat/conf.
154
DSE Search with Solr
Securing a Solr cluster
DSE Search data is completely or partially secured by using DataStax Enterprise security features.
DataStax Enterprise supports secure enterprise search using Apache Solr 4.3 and Lucene. The security
table summarizes the security features of DSE Search/Solr and other integrated components. DSE Search
data is completely or partially secured by using DataStax Enterprise security features:
•
Object permission management
•
Access to Solr documents, excluding cached data, can be limited to users who have been granted
access permissions. Permission management also secures tables used to store Solr data.
Transparent data encryption
•
Data at rest in Cassandra tables, excluding cached and Solr-indexed data, can be encrypted.
Encryption occurs on the Cassandra side and impacts performance slightly.
Client-to-node encryption
•
You can encrypt HTTP access to Solr data and internal, node-to-node Solr communication using SSL.
Enable SSL node-to-node encryption on the Solr node by setting encryption options in the dse.yaml
file as described in Client-to-node encryption.
Kerberos authentication
You can authenticate DSE Search users through Kerberos authentication using Simple and Protected
GSSAPI Negotiation Mechanism (SPNEGO). To use the SolrJ API against DSE Search clusters
with Kerberos authentication, client applications should use the SolrJ-Auth library and the DataStax
Enterprise SolrJ component as described in the solrj-auth-README.md file.
You can also use HTTP Basic Authentication, but this is not recommended.
HTTP Basic Authentication
When you enable Cassandra's internal authentication by specifying authenticator:
org.apache.cassandra.auth.PasswordAuthenticator in cassandra.yaml, clients must
use HTTP Basic Authentication to provide credentials to Solr services. Due to the stateless nature of
HTTP Basic Authentication, this can have a significant performance impact as the authentication process
must be executed on each HTTP request. For this reason, DataStax does not recommend using internal
authentication on DSE Search clusters in production. To secure DSE Search in production, enable
DataStax Enterprise Kerberos authentication.
To configure DSE Search to use Cassandra's internal authentication, follow this configuration procedure:
1. Comment AllowAllAuthenticator and uncomment the PasswordAuthenticator in cassandra.yaml to
enable HTTP Basic authentication for Solr.
#authenticator: org.apache.cassandra.auth.AllowAllAuthenticator
authenticator: org.apache.cassandra.auth.PasswordAuthenticator
#authenticator: com.datastax.bdp.cassandra.auth.PasswordAuthenticator
#authenticator: com.datastax.bdp.cassandra.auth.KerberosAuthenticator
2. Configure the replication strategy for the system_auth keyspace.
3. Start the server.
4. Open a browser, and go to the service web page. For example, assuming you ran the wikipedia demo,
go to http://localhost:8983/demos/wikipedia/.
The browser asks you for a Cassandra username and password.
Fast repair
Repair subranges of data in a cluster instead of running a nodetool repair operation on entire ranges.
Repairing subranges of data in a cluster is faster than running a nodetool repair operation on entire ranges
because all the data replicated during the nodetool repair operation has to be re-indexed. When you repair
a subrange of the data, less data has to be re-indexed.
To repair a subrange
155
DSE Search with Solr
Perform these steps as a rolling repair of the cluster, one node at a time.
1. Run the dsetool list_subranges command, using the approximate number of rows per subrange, the
beginning of the partition range (token), and the end of the partition range of the node.
dsetool list_subranges my_keyspace my_table 10000
113427455640312821154458202477256070485 0
The output lists the subranges.
Start Token
End Token
Estimated Size
--------------------------------------------------------------------------------------113427455640312821154458202477256070485
132425442795624521227151664615147681247 11264
132425442795624521227151664615147681247
151409576048389227347257997936583470460 11136
151409576048389227347257997936583470460 0
11264
2. Use the output of the previous step as input to the nodetool repair command.
nodetool repair my_keyspace my_table -st
113427455640312821154458202477256070485
-et 132425442795624521227151664615147681247
nodetool repair my_keyspace my_table -st
132425442795624521227151664615147681247
-et 151409576048389227347257997936583470460
nodetool repair my_keyspace my_table -st
151409576048389227347257997936583470460
-et 0
The anti-entropy node repair runs from the start to the end of the partition range.
Excluding hosts from Solr-distributed queries
You can exclude hosts from Solr-distributed queries.
You can exclude hosts from Solr-distributed queries by performing these steps on each node that you want
to send queries to.
1. Navigate to the Solr/conf directory:
• Packaged installs: /usr/share/dse/solr/conf
• Tarball installs: dse install/resources/solr/conf
2. Open the exclude.hosts file, and add the list of nodes to be excluded. Each name must be
separated by a newline character.
3. Update the list of routing endpoints on each node, by calling the JMX operation
refreshEndpoints() on the com.datastax.bdp:type=ShardRouter mbean.
Expiring a DSE Search column
Set a time when data expires to update a DSE Search column.
You can update a DSE Search column to set a time when data expires in these ways:
•
Configuring the high-performance update handler
•
•
Configuring per-document TTL causes removal of the entire document. Configuring per-field TTL
causes removal of the field only.
Using the Solr HTTP API
Using CQL to set TTL
If you configure TTL in the solrconfig.xml, and then use the Solr HTTP API or CQL to configure a different
TTL, the latter takes precedence.
Configuring expiration using the Solr HTTP API
156
DSE Search with Solr
Use the Solr HTTP API update command to set a time-to-live (TTL). You can construct a URL to update
data that includes the TTL per-document or per-field parameter:
•
Using the ttl parameter
Specifies per-document TTL. For example:
•
curl http://host:port/solr/mykeyspace.mytable/update?ttl=86400
Using the ttl.field name parameter
Specifies per-field TTL. For example:
curl http://host:port/solr/mykeyspace.mytable/update?ttl.myfield=86400
Configuring expiration using CQL
Using a CQL INSERT or UPDATE operation, you can set the TTL property. For example, continuing with
the example of using a collection set, insert a 5 minute (300 seconds) TTL property on the all columns of
the Einstein data:
INSERT INTO mysolr (id, name, title, body)
VALUES ('126', 'Albert Einstein', 'Success', 'If A is success
in life, then A equals x plus y plus z. Work is x; y is play;
and z is keeping your mouth shut.')
USING TTL 300;
After a few seconds, check the remaining time-to-live on the data:
SELECT TTL (name) FROM mykeyspace.mysolr WHERE id = '126';
The output after 9 seconds expire is:
ttl(name)
----------291
After the remaining time has passed, the data expires, and querying the data returns no results. If you
refresh the Solr Admin console, the number of documents is 3 instead of 4.
Configuring expiration scope
You can configure the solrconfig.xml to include the TTL per-document or per-field on data added
to the Solr index or Cassandra database. You construct a Solr HTTP API query to search the Solr index
using a ttl component. Depending on the configuration, TTL then applies to the entire document or just to a
named field.
To configure per-document or per-field TTL in the update handler:
1. Configure the high-performance update handler section of the solrconfig.xml.
•
For per-document TTL, add these lines to the high-performance updateHandler section:
<!-- The default high-performance update handler -->
<updateHandler class="solr.DirectUpdateHandler2">
. . .
•
<lst name="defaults">
<int name="ttl">1</int>
</lst>
For per-field TTL, add these lines to the updateHandler section:
<lst name = "defaults">
<int name = "ttl.<column/field name1>">1</int>
<int name = "ttl.<column/field name2>">1</int>
<int name = "ttl.<column/field name3>">1</int>
<int name = "ttl.<column/field name4>">1</int>
. . .
</lst>
2. Re-index the data by uploading the schema.xml and solrconfig.xml and reloading the Solr core.
157
DSE Search with Solr
Managing expired columns
After Cassandra expires a column using the time-to-live (TTL) mechanism, DSE Search/Solr can still find
the expired column. The column data remains in the index until one of the following conditions is met:
•
Re-indexing occurs due to a DSE Search ttl rebuild timeout.
•
Set the ttl rebuild timeout properties in the dse.yaml file.
All columns in a row expire due to the Cassandra time-to-live (TTL) mechanism, triggering removal of
the entire row/Solr document from the index.
Setting the ttl rebuild timeout properties is the recommended method for managing expired columns.
Changing the HTTP interface to Apache JServe Protocol
Configure search to use the AJP (Apache JServe Protocol). This capability is typically used where https
serves a web application and DSE Search powers the backend.
In addition to the widely-used HTTP interface, you can configure DSE search to use the AJP (Apache
JServe Protocol). AJP is an optimized, binary version of HTTP that facilitates Tomcat communication with
an Apache web server using mod_jk. This capability is typically used where https serves a web application
and DSE Search powers the backend. By default the AJP connector is disabled. To enable the AJP
connector, uncomment the connector configuration in the Tomcat server.xml file. For example, remove the
comments as follows:
<!-- Define an AJP 1.3 Connector on port 8009 -->
Connector port="8009" protocol="AJP/1.3" redirectPort="8443"
The Tomcat server.xml file is located in the following directory:
•
•
Packaged installs: /usr/share/dse/tomcat/conf/
Tarball installs: install_location/resources/tomcat/conf/
Shard transport options for DSE Search/Solr communications
A custom, TCP-based communications layer for Solr is the default type in DataStax Enterprise. To improve
Solr inter-node communications and avoid distributed deadlock during queries, switch from the HTTPbased communications to the netty non-blocking communications layer.
Available in DataStax Enterprise 4.0 is a custom, TCP-based communications layer for Solr. The new
communications layer is an alternative to the default HTTP-based, Tomcat-backed interface, which
customers say is slow and resource intensive. The new communications layer improves Solr inter-node
communications in several ways:
•
•
•
•
Lowers latency
Reduces resource consumption
Increases throughput even while handling thousands of concurrent requests
Provides nonblocking I/O processing
To avoid distributed deadlock during queries, switch from the HTTP-based communications to the new
non-blocking communications layer.
The TCP-based communications layer for Solr supports client-to-node and node-to-node encryption using
SSL, but does not support Kerberos.
Configure the shard transport options in the dse.yaml file to select HTTP- or TCP-based communication.
The dse.yaml file is located in the following directories:
•
•
Packaged installs: /etc/dse/cassandra
Tarball installs: install_location/resources/dse/conf
The shard_transport_options in the dse.yaml file for managing inter-node communication between Solr
nodes are:
•
158
type: http or netty
DSE Search with Solr
•
The default type, http, configures plain old Solr communication that uses the standard HTTP-based
communications interface. Choosing the netty type configures TCP-based Solr communications. If you
chose netty, the following netty options are applicable.
netty_server_port: 8984
•
The TCP listen port, mandatory if use the netty type, or if you want to migrate to the netty type from the
http type later. If you plan to use the http type indefinitely, either comment netty_server_port or set it to
-1.
netty_server_acceptor_threads
•
The number of server acceptor threads. The default is number of available processors.
netty_server_worker_threads
•
The number of server worker threads. The default is number of available processors times 8.
netty_client_worker_threads
•
The number of client worker threads. The default is number of available processors times 8.
netty_client_max_connections
•
The maximum number of client connections. The default is 100.
netty_client_request_timeout
The client request timeout in milliseconds. The default is 60000.
Tuning DSE Search performance
Topics for performance tuning and solving performance degradation, high memory consumption, or other
problem swith DataStax Enterprise Search nodes.
About this task
In the event of a performance degradation, high memory consumption, or other problem with DataStax
Enterprise Search nodes, try:
•
•
•
•
•
•
•
•
•
•
Using Cassandra table compression
Configuring the Search Handler
Configuring the update handler and autoSoftCommit
Changing the stack size and memtable space
Managing the data consistency level
Configuring the available indexing threads
Managing caching
Adding replicas to increase read performance
Changing the replication factor
Configuring re-indexing and repair
Commit and query latency MBeans
To troubleshoot, tune performance, and resolve consistency issues, use commit, merge, query, and update
metrics MBeans.
DataStax Enterprise 4.0 provides commit and query metrics MBeans for troubleshooting and tuning
performance and consistency issues.
The following path identifies the commit metrics and query metrics mbeans:
type=search,index=<core>,name=CommitMetrics
type=search,index=<core>,name=QueryMetrics
<core> is the name of the Solr core referenced by the metric.
159
DSE Search with Solr
For example, the following figure shows the path to CommitMetrics and QueryMetrics for the demo.solr
core in jconsole.
Commit metrics MBean
The commit metrics MBean is useful for troubleshooting index performance and resolving data consistency
issues that are caused by asynchronous commits between different index replicas.
The commit metrics MBean is useful for troubleshooting index performance as well as data consistency
issues caused by asynchronous commits between different index replicas. Using this mbean is also useful
for fine-tuning indexing backpressure. Back pressure is pausing or blocking the buffering of incoming
requests after reaching a threshold until internal processing of buffered requests catches up.
The commit metrics mbean records the amount of time spent to execute two main phases of a commit
operation on the index.
Main operational phases
The main phases of a commit operation on the index are:
•
•
FLUSH: comprising the time spent by flushing the async indexing queue.
EXECUTE: comprising the time spent by actually executing the commit on the index.
Commit metrics mbean operations
The commit metrics mbean measures latency in microseconds. Operations are:
160
•
setEnabled(boolean enabled)
•
Enables/disables metrics recording. Enabled by default.
isEnabled()
•
Checks that metrics recording is enabled.
getLatencyPercentile(String phase, double percentile)
DSE Search with Solr
•
Gets a commit latency percentile by its phase.
getRecordedLatencyCount(String phase)
•
Gets the total count of recorded latency metrics by its commit phase.
getUnrecordedLatencyCount()
•
Gets the total count of unrecorded latency values due to exceeding the maximum tracked latency,
which is 10 minutes.
resetLatency(String phase)
•
Resets latency metrics for the given commit phase.
resetLatencies()
Resets all latency metrics.
Commit metrics mbean operations use the FLUSH and EXECUTE commit phase names.
Query metrics MBean
The query metrics MBean is useful for troubleshooting query performance, tuning the Solr configuration,
such as the schema and caches, and tuning server resources, such as the JVM heap.
About this task
The query metrics MBean is useful for troubleshooting query performance, tuning the Solr configuration,
such as the schema and caches, and tuning server resources, such as the JVM heap. The query
metricsMBean records the amount of time spent to execute several main phases of distributed query on
the index.
Main operational phases
The main phases of a distributed query operation are:
•
COORDINATE
•
Comprises the total amount of time spent by the coordinator node to distribute the query and gather/
process results from shards. This value is computed only on query coordinator nodes.
EXECUTE
•
Comprises the time spent by a single shard to execute the actual index query. This value is computed
on the local node executing the shard query.
RETRIEVE
Comprises the time spent by a single shard to retrieve the actual data from Cassandra. This value will
be computed on the local node hosting the requested data.
Query metrics mbean operations
The query metrics mbean measures latency in microseconds. Metrics can be grouped by query, by
providing an additional query.name parameter. For example, assuming you are using a Solr core named
demo.solr and have indexed a field named type, this URL provides the additional query.name parameter:
http://localhost:8983/solr/demo.solr/select/?q=type:1&query.name=myquery
All metrics collected under a given query name are recorded and retrieved separately, as shown in the
following list of operations. If no query name is provided, all metrics are recorded together.
Operations are:
•
setEnabled(boolean enabled)
•
Enables/disables metrics recording. Enabled by default.
isEnabled()
161
DSE Search with Solr
•
Checks if metrics recording is enabled.
getLatencyPercentile(String phase, String query, double percentile)
•
Gets a query latency percentile by its query name, which is optional and can be null, and phase.
getRecordedLatencyCount(String phase, String query)
•
Gets the total count of recorded latency metrics by its query name, which is optional and can be null,
and phase.
getUnrecordedLatencyCount()
•
Gets the total count of unrecorded latency values due to exceeding the maximum tracked latency,
which is 10 minutes.
resetLatency(String query)
•
Resets latency metrics for the given query name, which is optional and can be null.
resetLatencies()
Resets all latency metrics.
Query metrics mbean operations use the phase names previously listed.
Using mbeans to evaluate performance
The following example shows how to use the mbeans on Linux to obtain information about performance
while running the DataStax Solr stress test demo.
1. Start a single Solr node.
2. Start jconsole using the PID of the Solr node: For example:
sudo jconsole 1284
3. On Linux, for example, execute these scripts to run the Solr stress demo in dse-4.0.0/demos/
solr_stress.
./1-add-schema.sh
./2-run-benchmark.sh --clients=10 --loops=10000 --type=both
The demo creates a Solr core named demo.solr and indexes 50,000 documents.
4. In jconsole, expand com.datastax.bdp > search > demo.solr.
The CommitMetrics and QueryMetrics mbean items appear.
5. In jconsole, in Search > demo.solr > CommitMetrics > Operations > getLatencyPercentile, type
EXECUTE in the p0 text entry box and 0.95 in the p1 text entry box. Click the getLatencyPercentile
button.
The Operation return value, 582 microseconds, appears:
162
DSE Search with Solr
6. Click OK.
7. Query Solr 20,000 times using the query.name parameter. For example:
curl "http://localhost:8983/solr/demo.solr/select/?
q=type:1&query.name=myquery"
curl "http://localhost:8983/solr/demo.solr/select/?
q=type:3&query.name=myquery"
8. In jconsole, in Search > demo.solr > QueryMetrics Operations getLatencyPercentile, type
EXECUTE in the p0 text entry box, myquery in the p1 text entry box, and 95.0 in the P2 text entry box.
The Operation return value, 97 microseconds, appears.
163
DSE Search with Solr
Using table compression
Configure data compression on a per-table basis to optimize performance of read-dominated tasks.
Search nodes typically engage in read-dominated tasks, so maximizing storage capacity of nodes,
reducing the volume of data on disk, and limiting disk I/O can improve performance. In Cassandra 1.0 and
later, you can configure data compression on a per-table basis to optimize performance of read-dominated
tasks.
Configuration affects the compression algorithm for compressing SSTable files. For read-heavy
workloads, such as those carried by Enterprise Search, LZ4 compression is recommended.
Compression using the LZ4 compressor is enabled by default when you create a table. You can change
compression options using CQL. Developers can also implement custom compression classes using the
org.apache.cassandra.io.compress.ICompressor interface. You can configure the compression chunk size
for read/write access patterns and the average size of rows in the table.
Configuring the update handler and autoSoftCommit
To configure the update handler, set the default high-performance update handler flag.
You need to configure the solrconfig.xml to use near real-time capabilities in Solr by setting the
default high-performance update handler flag.
For example, the Solr configuration file for the Wikipedia demo sets this flag as follows and uncomments
the autoSoftCommit element:
<!-- The default high-performance update handler -->
<updateHandler class="solr.DirectUpdateHandler2">
. . .
<autoSoftCommit>
<maxTime>1000</maxTime>
</autoSoftCommit>
</updateHandler>
164
DSE Search with Solr
The autoCommit element has been removed to prevent hard commits that hit the disk and flush the cache.
The soft commit forces uncommitted documents into internal memory. When data is committed, is it
immediately available after the commit.
The autoSoftCommit element uses the maxTime update handler attribute. The update handler attributes
enable near real-time performance and trigger a soft commit of data automatically, so checking
synchronization of data to disk is not necessary. This table describes both update handler options.
Attribute
Default
Description
maxDocs
No default
Maximum number of documents to add since the last soft
commit before automatically triggering a new soft commit
maxTime
1000
Maximum expired time in milliseconds between the addition of a
document and a new, automatically triggered soft commit
For more information about the update handler and modifying solrconfig.xml, see the Solr
documentation.
Configuring update performance
If updates take too long and you changed the default autoSoftCommit from the default 1000 ms to a higher
value, reset autoSoftCommit in the solrconfig.xml to its default value.
Changing the stack size and memtable space
Increasing the stack size can improve performance under Tomcat.
Some Solr users have reported that increasing the stack size improves performance under Tomcat. To
increase the stack size, uncomment and modify the default -Xss128k setting in the cassandra-env.sh
file. Also, decreasing the memtable space to make room for Solr caches might improve performance.
Modify the memtable space using the memtable_total_space_in_mb property in the cassandra.yaml file.
Managing the consistency level
Configure how up-to-date and synchronized a row of data is on all of its replicas.
Consistency refers to how up-to-date and synchronized a row of data is on all of its replicas. Like
Cassandra, DSE-Search extends Solr by adding an HTTP parameter, cl, that you can send with Solr data
to tune consistency. The format of the URL is:
curl "http://host:port/solr/keyspace.table/update?cl=ONE"
The cl parameter specifies the consistency level of the write in Cassandra on the client side. The
default consistency level is QUORUM, but you can change the default globally, on the server side using
Cassandra’s drivers and client libraries.
Setting the consistency level using SolrJ
SolrJ does not allow setting the consistency level parameter using a Solr update request. To set the
consistency level parameter:
HttpSolrServer httpSolrServer = new HttpSolrServer ( url );
httpSolrServer . getInvariantParams (). add ( "cl" , "ALL" );
For more information, see the Data Consistency in DSE Search blog.
Configuring the available indexing threads
Improve performance on machines that have multiple CPU cores.
DSE Search provides a multi-threaded indexing implementation to improve performance on machines
having multiple CPU cores. All index updates are internally dispatched to a per-CPU core indexing
thread pool and executed asynchronously. This implementation allows for greater concurrency and
165
DSE Search with Solr
parallelism, but as a consequence, index requests return a response before the indexing operation
is actually executed. The number of available indexing threads per Solr core is by default equal
to number of available CPU cores times 2. The available threads can be configured by editing the
max_solr_concurrency_per_core parameter in the dse.yaml configuration file; if set to 1, DSE Search
uses the legacy synchronous indexing implementation.
Also, DSE Search provides advanced, JMX-based, configurability and visibility through the IndexPool-ks.cf
(where ks.cf is the name of a DSE Search Solr core) MBean under the com.datastax.bdp namespace.
Managing caching
Steps for managing cache depending on version.
About this task
DataStax Enterprise defaults to using NRTCachingDirectoryFactory, which is the OS Solr-recommended
setting for real-time performance. These default properties specify where files are cached and files are
managed:
•
•
maxMergeSizeMB = 4.0 MB
maxCachedMB = 48.0 MB
Typically, when using DataStax Enterprise, you change the factory class to
DSENRTCachingDirectoryFactory and configure the file properties in the solrconfig.xml.
Procedure
Configuring where files are cached:
1. Open solrconfig.xml for editing.
2. Add a directoryFactory element to solrconfig.xml of type DSENRTCachingDirectoryFactory. For
example:
<directoryFactory name="DirectoryFactory"
class="com.datastax.bdp.cassandra.index.solr.DSENRTCachingDirectoryFactory">
<double name="maxmergesizemb">5.0</double>
<double name="maxcachedmb">32.0</double>
</directoryFactory>
3. Adjust the DirectoryFactory attributes, using lowercase property names, for your environment.
•
maxmergesizemb
•
The threshold (MB) for writing a merge segment to memory or to the file system. If the estimated
size of merging a segment is less than maxmergesizemb, the merge segment is written to the
memory; otherwise, it is written to the file system.
maxcachemb
The maximum value (MB) for writing a merged segment into memory as opposed to disk.
Tuning index size and range query speed
You can trade off Solr index size for range query speed and vice versa. You make this tradeoff to suit a
particular use case and on a core-by-core basis.
In DataStax Enterprise, you can trade off Solr index size for range query speed and vice versa. You make
this tradeoff to suit a particular use case and on a core-by-core basis by setting up the precision step of two
special token field types used by DataStax Enterprise.
Use extreme care when performing this tuning. This advanced tuning feature is recommended for use in
rare cases. In most cases, using the defaults is the best. To perform this tuning, you change the precision
step of one or both DataStax Enterprise internal field types:
•
166
token_long
DSE Search with Solr
•
Used for filtering over token fields during query routing.
ttl_long
Used for searching for expiring documents.
Change the precision step as follows:
1. In the fieldType definition, set the class attribute of token_long and ttl_long to
solr.TrieLongField.
2. Set the precisionStep attribute from the default 8 to another number. Choose this number based
on an understanding of its impact. Usually, a smaller precision step increases the index size and range
query speed, while a larger precision step reduces index size, but potentially reduces range query
speed.
The following snippet of the schema.xml shows an example of the required configuration of both field
types:
<?xml version="1.0" encoding="UTF-8" ?>
<schema name="test" version="1.0">
<types>
. . .
<fieldType name="token_long" class="solr.TrieLongField"
precisionStep="16" />
<fieldType name="ttl_long" class="solr.TrieLongField" precisionStep="16" />
. . .
</types>
<fields>
. . .
</schema>
DataStax Enterprise ignores one or both of these field type definitions and uses the default precision step if
you make any of these mistakes:
•
•
•
The field type is defined using a name other than token_long or ttl_long.
The class is something other than solr.TrieLongField.
The precision step value is not a number. DataStax Enterprise logs a warning.
The definition of a fieldType alone sets up the special field. You do not need to use token_long or
ttl_long types as fields in the <fields> tag.
Increasing read performance by adding replicas
Increase DSE Search read performance by configuring replicas.
About this task
You can increase DSE Search read performance by configuring replicas just as you do in Cassandra. You
define a strategy class, the names of your data centers, and the number of replicas you want. For example,
you can add replicas using the NetworkToplogyStrategy replica placement strategy. To configure this
strategy, you can use CQL.
Procedure
For example, if you are using a PropertyFileSnitch, perform these steps:
1. Check the data center names of your nodes using the nodetool command.
./nodetool -h localhost ring
The data center names, DC1 and DC2 in this example, must match the data center name configured for
your snitch.
2. Start CQL on the DSE command line and create a keyspace that specifies the number of replicas you
want.
167
DSE Search with Solr
Set the number of replicas in data centers, one replica in data center 1 and three in data center 2. For
more information about adding replicas, see Choosing Keyspace Replication Options.
Changing the replication factor for a Solr keyspace
Steps for changing the keyspace replication factor after the solrconfig.xml and schema.xml files are posted.
About this task
This example assumes the solrconfig.xml and schema.xml files have already been posted using
mykeyspace.mysolr in the URL, which creates a keyspace named mykeyspace that has a default
replication factor of 1. You want three replicas of the keyspace in the cluster, so you need to update the
keyspace replication factor.
Procedure
To change the keyspace replication factor
1. Check the name of the data center of the Solr/Search nodes.
./nodetool -h localhost ring
The output tells you that the name of the data center for your node is, for example, datacenter1.
2. Use CQL to change the replication factor of the keyspace from 1 to 3.
ALTER KEYSPACE mykeyspace WITH REPLICATION = { 'class' :
'NetworkTopologyStrategy', 'datacenter1' : 3 };
If you have data in a keyspace and then change the replication factor, run nodetool repair to avoid
having missing data problems or data unavailable exceptions.
Configuring re-indexing
Change the size of the RAM buffer and increase the soft commit time in the solrconfig.xml file to tune the
performance of re-indexing and index building.
About this task
When running the RELOAD command using the reindex or deleteAll options, a long delay might indicate
that tuning is needed. Tune the performance of re-indexing and index rebuilding by making a few changes
in the solrconfig.xml file.
Procedure
1. Increase the size of the RAM buffer, which is set to 100MB by default, to 125 for example.
<indexConfig>
<useCompoundFile>false</useCompoundFile>
<ramBufferSizeMB>125</ramBufferSizeMB>
<mergeFactor>10</mergeFactor>
. . .
2. Increase the soft commit time, which is set to 1000 ms by default, to a larger value. For example,
increase the time to 15-16 minutes:
<autoSoftCommit>
<maxTime>1000000</maxTime>
</autoSoftCommit>
The downside of changing the autoSoftCommit attribute is that newly updated rows take longer than
usual (1000 ms) to appear in search results.
168
DSE Search with Solr
DSE Search/Solr versus Open Source Solr
A comparison of DSE Search and Open Source Solr.
By virtue of its integration into DataStax Enterprise, differences exist between DSE Search/Solr and Open
Source Solr (OSS).
Major differences
The major differences in capabilities are:
Capability
DSE
OS Solr
Description
Includes a database
yes
no
A user has to create an interface to add a database
to OSS.
Indexes real-time data
yes
no
Cassandra ingests real-time data and Solr indexes
the data.
Provides an intuitive way yes
to update data
no
DataStax provides a SQL-like language and
command-line shell, CQL, for loading and updating
data. Data added to Cassandra shows up in Solr
Indexes Hadoop output
without ETL
yes
no
Cassandra ingests the data, Solr indexes the data,
and you run MapReduce against that data in one
cluster.
Supports data
distribution
yes
yes [1]
DataStax Enterprise distributes Cassandra realtime, Hadoop, and Solr data to multiple nodes in a
cluster transparently.
Balances loads on
nodes/shards
yes
no
Unlike OSS and Solr Cloud loads can be rebalanced
efficiently.
Spans indexes over
multiple data centers
yes
no
A cluster can have more than one data center for
different types of nodes.
Automatically re-indexes yes
Solr data
no
The only way to re-index data in OSS is to have the
client re-ingest everything.
Stores data added
through Solr in
Cassandra
yes
no
Data updated using the Solr API shows up in
Cassandra.
Makes durable updates
to data
yes
no
Updates are durable and written to the Cassandra
commit log regardless of how the update is made.
Upgrades of Lucene
preserve data
yes
no
DataStax integrates Lucene upgrades periodically
and when you upgrade DSE, data is preserved.
OSS users must re-ingest all their data when
upgrading to Lucene.
Security
yes
no
DataStax has extended SolrJ to protect internal
communication and HTTP access. Solr data can be
encrypted and audited.
[1] Requires using Zookeeper.
Minor differences
Minor differences between DSE Search and OSS include:
169
DSE Search with Solr
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
The methods for importing data using the Data Import Handler differ. The DSE Search/Solr method is
configuration-driven.
You launch DSE Search by starting a DataStax Enterprise node in DSE Search mode. You start Solr
using java -jar start.jar.
DSE Search terminology used to describe objects differs from OSS terminology. The Defining key Solr
terms section lists the differences.
Delete by query in DSE Search differs from OSS. Deletions begin immediately. You do not need to post
a commit after posting the delete command.
The process for creating an index and reloading a schema differs.
DSE Search has removed the Optimize button from the Core Admin UI.
In the DSE Search schema, if you do not configure the uniqueKey field as stored (stored="true"),
DataStax Enterprise forces that flag to be true.
This change is necessary to make distributed search work.
Behavior differs between DSE Search and OSS when you configure a non-unique field as not stored.
In OSS, the data is lost, whereas in DSE Search, the data is stored in Cassandra. The field does not
show up in the search results of DSE Search or OSS.
DataStax provides a real-time caching directory factory flag, DSENRTCachingDirectoryFactory, that
you can use to specify where files are cached.
The autoCommit element in the Solrconfig.xml is removed in DSE Search/Solr and the autoSoftCommit
element is uncommented. You do not need to use the commit attribute in a Solr HTTP update
command.
In OSS the autoCommit element is present and uncommented. The autoSoftCommit is commented out.
You need to use the commit attribute in a Solr HTTP update command.
OSS supports relative paths set by the <lib> property in the solrconfig.xml, but DSE Search/
Solr does not. Configuring the Solr library path describes a workaround for this issue that DataStax
Enterprise will address in a future release.
When using dynamic fields, the default column limit controls the maximum number of indexed columns
(overall, not just dynamic field ones), as well as columns returned during queries. The column limit
prevents out of memory errors caused by using too many dynamic fields.
DSE Search/Solr uses the Cassandra commit log instead of the Solr updateLog. Atomic updates and
real-time get are not supported. In Cassandra, a write is atomic at the row-level.
In OSS, you set the data location for saving Solr data files by configuring the <dataDir>xxx</dataDir>
element in the solrconfig.xml. DSE Search ignores this element and saves data in Cassandra
data directory/solr.data unless you specify another location during start up.
DSE Search/Solr supports a custom version of the Solr join operation.
Pseudo join and pivot faceting, not fully supported by DataStax Enterprise, do not belong in the differences
list because OSS does not support these, or any other OSS features, in distributed mode. OSS does not
distribute data in a scalable, peer-to-peer system like DataStax Enterprise does.
Request processing and data transformation
Use the custom update request processor (URP) to extend the Solr URP. Use the field input/output
transformer API as an option to the input/output transformer support in OS Solr.
DataStax Enterprise supports a field input/output tranformer API, the classic Solr update request processor
(URP), and a custom URP chain for processing requests and transforming data. The DataStax Enterprise
URP implementation provides similiar functionality to the Solr URP chain, but appears as a plugin to Solr.
The classic URP is invoked when updating a document using HTTP, the custom URP when updating a
table using Cassandra. If both classic and custom URPs are configured, the classic version is executed
first. A field input/output transformer, an alternative for handling update requests, is executed later than a
URP at indexing time.
170
DSE Search with Solr
To configure the custom URP, define the following element in the solrconfig.xml:
<dseUpdateRequestProcessorChain name="dse">
<processor
class="com.datastax.bdp.search.solr.functional.DSEUpdateRequestProcessorFactoryExample">
</processor>
</dseUpdateRequestProcessorChain>
API for transforming Cassandra/Solr data
Use the field input/output transformer API to the input/output transformer support in OS Solr.
About this task
DSE Search/Solr includes the released version of a plugin API for Solr updates and a plugin to the
CassandraDocumentReader. The plugin API transforms data from the secondary indexing API before
data is submitted to Solr. The plugin to the CassandraDocumentReader transforms the results data from
Cassandra to Solr.
Using the API, applications can tweak a Solr Document before it is mapped and indexed according to the
schema.xml. The API is a counterpart to the input/output transformer support in OS Solr.
The field input transformer (FIT) requires a trailing Z for date field values.
Procedure
To use the API:
1. Define the plugin in the solrconfig.xml for a Cassandra table (Solr Core).
<fieldInputTransformer name="dse" class="
com.datastax.bdp.cassandra.index.solr.functional.
BinaryFieldInputTransformer">
</fieldInputTransformer>
<fieldOutputTransformer name="dse" class="
com.datastax.bdp.cassandra.index.solr.functional.
BinaryFieldOutputTransformer">
</fieldOutputTransformer>
2. Write a transformer class something like this reference implementation to tweak the data in some way.
3. Export the class to a JAR, and place the JAR in this location:
•
•
Tarball installs: install-location/resources/solr/lib
Packaged installs: /usr/share/dse/solr/lib
The JAR is added to the CLASSPATH automatically.
4. Test your implementation using something like the reference implementation.
FIT reference implementation
Field input and output transformer (FIT) class examples.
About this task
Here are examples of field input and output transformer (FIT) classes.
Input transformer example
package com.datastax.bdp.cassandra.index.solr.functional;
import java.io.IOException;
import org.apache.commons.codec.binary.Hex;
import org.apache.commons.lang.StringUtils;
171
DSE Search with Solr
import org.apache.lucene.document.Document;
import org.apache.solr.core.SolrCore;
import org.apache.solr.schema.SchemaField;
import com.datastax.bdp.cassandra.index.solr.FieldInputTransformer;
import org.apache.solr.schema.IndexSchema;
public class BinaryFieldInputTransformer extends FieldInputTransformer
{
@Override
public boolean evaluate(String field)
{
return field.equals("binary");
}
@Override
public void addFieldToDocument(SolrCore core,
IndexSchema schema,
String key,
Document doc,
SchemaField fieldInfo,
String fieldValue,
float boost,
DocumentHelper helper)
throws IOException
{
try
{
byte[] raw = Hex.decodeHex(fieldValue.toCharArray());
byte[] decomp = DSP1493Test.decompress(raw);
String str = new String(decomp, "UTF-8");
String[] arr = StringUtils.split(str, ",");
String binary_name = arr[0];
String binary_type = arr[1];
String binary_title = arr[2];
SchemaField binaryNameField =
core.getSchema().getFieldOrNull("binary_name");
SchemaField binaryTypeField =
core.getSchema().getFieldOrNull("binary_type");
SchemaField binaryTitleField =
core.getSchema().getFieldOrNull("binary_title");
helper.addFieldToDocument(core, core.getSchema(), key, doc,
binaryNameField, binary_name, boost);
helper.addFieldToDocument(core, core.getSchema(), key, doc,
binaryTypeField, binary_type, boost);
helper.addFieldToDocument(core, core.getSchema(), key, doc,
binaryTitleField, binary_title, boost);
}
catch (Exception ex)
{
throw new RuntimeException(ex);
}
}
}
Output transformer example
package com.datastax.bdp.cassandra.index.solr.functional;
import java.io.IOException;
import org.apache.commons.lang.StringUtils;
import org.apache.lucene.index.FieldInfo;
172
DSE Search with Solr
import org.apache.lucene.index.StoredFieldVisitor;
import com.datastax.bdp.cassandra.index.solr.FieldOutputTransformer;
public class BinaryFieldOutputTransformer extends FieldOutputTransformer
{
@Override
public void binaryField(FieldInfo fieldInfo, byte[] value,
StoredFieldVisitor visitor, DocumentHelper helper) throws
IOException
{
byte[] bytes = DSP1493Test.decompress(value);
String str = new String(bytes, "UTF-8");
String[] arr = StringUtils.split(str, ",");
String binary_name = arr[0];
String binary_type = arr[1];
String binary_title = arr[2];
FieldInfo binary_name_fi = helper.getFieldInfo("binary_name");
FieldInfo binary_type_fi = helper.getFieldInfo("binary_type");
FieldInfo binary_title_fi = helper.getFieldInfo("binary_title");
visitor.stringField(binary_name_fi, binary_name);
visitor.stringField(binary_type_fi, binary_type);
visitor.stringField(binary_title_fi, binary_title);
}
}
Interface for custom field types
The CustomFieldType interface marks Solr custom field types and provides their actual stored field type.
About this task
DataStax Enterprise 4.0 implements a CustomFieldType interface that marks Solr custom field types
and provides their actual stored field type. The custom field type stores an integer trie field as a string
representing a comma separated list of integer values: when indexed, the string is split into its integer
values, each one indexed as a trie integer field. This class effectively implements a multi-valued field based
on its string representation.
To use the CustomFieldType interface:
1. Implement a custom field type class something like the following reference implementation.
2. Export the class to a JAR, and place the JAR in this location:
•
•
Packaged installs: usr/share/dse
Tarball installs: install_location/resources/dse/lib
The JAR is added to the CLASSPATH automatically.
Reference implementation
Here is an example of a custom field type class:
package com.datastax.bdp.search.solr.functional;
import
import
import
import
import
import
import
import
com.datastax.bdp.search.solr.CustomFieldType;
java.util.ArrayList;
java.util.List;
org.apache.lucene.index.IndexableField;
org.apache.solr.schema.FieldType;
org.apache.solr.schema.SchemaField;
org.apache.solr.schema.StrField;
org.apache.solr.schema.TrieField;
public class CustomTestField extends TrieField implements CustomFieldType
173
DSE Search with Solr
{
public CustomTestField()
{
this.type = TrieField.TrieTypes.INTEGER;
}
@Override
public FieldType getStoredFieldType()
{
return new StrField();
}
@Override
public boolean multiValuedFieldCache()
{
return true;
}
@Override
public List<IndexableField> createFields(SchemaField sf, Object value,
float boost)
{
String[] values = ((String) value).split(" ");
List<IndexableField> fields = new ArrayList<IndexableField>();
for (String v : values)
{
fields.add(createField(sf, v, boost));
}
return fields;
}
@Override
public String toInternal(String value)
{
return value;
}
@Override
public String toExternal(IndexableField f)
{
return f.stringValue();
}
}
174
Deploying
Deploying
Deployment topics.
Production deployment planning
Production deployment planning requires knowledge of the initial volume of data to store and an estimate
of the typical application workload.
The Cassandra 2.0 topic Planning a cluster deployment provides guidance for planning a DataStax
Enterprise cluster. The following resources and guidelines are also recommended:
•
•
•
•
•
•
The DataStax Enterprise Reference Architecture white paper.
For EC2 deployments, see:
• Installing a DSE cluster on EC2
• What is the story with AWS storage
• Get in the Ring with Cassandra and EC2
DataStax Enterprise requires a solid network layer. Although not required, jumbo frames are
recommended to improve streaming performance during processes such as bootstrapping and repair.
Hadoop and Solr nodes require their own nodes/disks and have specific hardware requirements.
See the Hadoop and Solr documentation for more information when determining your capacity
requirements.
DataStax neither supports nor recommends using Network Attached Storage (NAS) because of
performances issues, such as network saturation, I/O overload, pending-task swamp, excessive
memory usage, and disk contention.
If using a firewall, make sure that nodes within a cluster can reach each other. See Configuring firewall
port access.
Configuring replication
Choose a data partitioner and replica placement strategy.
About this task
Cassandra performs replication to store multiple copies of data on multiple nodes for reliability and fault
tolerance. To configure replication, you need to choose a data partitioner and replica placement strategy.
Data partitioning determines how data is placed across the nodes in the cluster. For information about how
this works, see Data distribution and replication. Nodes communicate with each other about replication
and other things using the gossip protocol. Be sure to configure gossip, as described in About internode
communications (gossip).
Virtual nodes
Virtual nodes simplify many tasks in Cassandra, such as eliminating the need to determine the partition
range (calculate and assign tokens), rebalancing the cluster when adding or removing nodes, and
replacing dead nodes. For a complete description of virtual nodes and how they work, see About virtual
nodes, and the Virtual nodes in Cassandra 1.2 blog.
Attention: DataStax Enterprise 4.0 turns off virtual nodes (vnodes) by default. DataStax does
not recommend turning on vnodes for Hadoop or Solr nodes, but you can use vnodes for any
Cassandra-only cluster, or a Cassandra-only data center in a mixed Hadoop/Solr/Cassandra
deployment.
175
Deploying
Using virtual nodes
In the cassandra.yaml, file uncomment num_tokens and leave the initial_token parameter unset.
Guidelines for using virtual nodes include:
•
Determining the num_tokens value
•
The initial recommended value for num_tokens is 256. For more guidance, see Setting up virtual nodes.
Migrating existing clusters
•
To upgrade existing clusters to virtual nodes, see Enabling virtual nodes on an existing production
cluster.
Using a mixed architecture
Cassandra supports using virtual node-enabled and non-virtual node data centers. For example, a
single cluster could have a cassandra only data center with vnodes enabled and a search data center
without vnodes.
Disabling virtual nodes
To disable virtual nodes:
1. In the cassandra.yaml file, set num_tokens to 1.
num_tokens: 1
2. Uncomment the initial_token property and set it to 1 or to the value of a generated token for a multinode cluster.
Using the single-token-per-node architecture in DSE 3.1 and above
If you don't use virtual nodes, you must make sure that each node is responsible for roughly an equal
amount of data. To do this, assign each node an initial-token value and calculate the tokens for each
data center as described in Generating tokens located in the Datastax Enterprise 3.0 documentation. You
can also use the Murmur3Partitioner and calculate the tokens as described in Cassandra 1.2 Generating
tokens.
Partitioner settings
You can use either the Murmur3Partitioner or RandomPartitioner with virtual nodes.
The Murmur3Partitioner (org.apache.cassandra.dht.Murmur3Partitioner) is the default
partitioning strategy for new Cassandra clusters (1.2 and above) and the right choice for new clusters in
almost all cases. You can only use Murmur3Partitioner for new clusters; you cannot change the partitioner
in existing clusters. If you are switching to the 1.2 cassandra.yaml, be sure to change the partitioner
setting to match the previous partitioner.
The RandomPartitioner (org.apache.cassandra.dht.RandomPartitioner) was the default
partitioner prior to Cassandra 1.2. You can continue to use this partitioner when migrating to virtual nodes.
Snitch settings
A snitch determines which data centers and racks are written to and read from. It informs Cassandra about
the network topology so that requests are routed efficiently and allows Cassandra to distribute replicas by
grouping machines into data centers and racks. All nodes must have exactly the same snitch configuration.
The following sections describe three commonly-used snitches. All available snitches
are described in the Cassandra documentation. The default endpoint_snitch is the
DseDelegateSnitch. The default snitch delegated by this snitch is the DseSimpleSnitch
(org.apache.cassandra.locator.DseSimpleSnitch). You set the snitch used by the
DseDelegateSnitch in the dse.yaml file:
•
176
Packaged installs: /etc/dse/dse.yaml
Deploying
•
Tarball installs: install_location/resources/dse/conf/dse.yaml
DseSimpleSnitch
Use DseSimpleSnitch only for development in DataStax Enterprise deployments. This snitch logically
configures each type of node in separate data centers to segregate the analytics, real-time, and search
workloads.
When defining your keyspace, use Analytics, Cassandra, or Search for your data center names.
SimpleSnitch
For a single data center (or single node) cluster, the SimpleSnitch is usually sufficient. However, if you plan
to expand your cluster at a later time to multiple racks and data centers, it is easier if you use a rack and
data center aware snitch from the start, such as the RackInferringSnitch.
PropertyFileSnitch
The PropertyFileSnitch allows you to define your data center and rack names to be whatever you want.
Using this snitch requires that you define network details for each node in the cluster in the cassandratopology.properties configuration file.
•
•
Packaged installs: /etc/dse/cassandra/cassandra-topology.properties
Tarball installs: install_location/resources/cassandra/conf/cassandratopology.properties
Every node in the cluster should be described in this file, and specified exactly the same on every node in
the cluster.
For example, suppose you had non-uniform IPs and two physical data centers with two racks in each, and
a third logical data center for replicating analytics data, you would specify them as follows:
# Data Center One
175.56.12.105=DC1:RAC1
175.50.13.200=DC1:RAC1
175.54.35.197=DC1:RAC1
120.53.24.101=DC1:RAC2
120.55.16.200=DC1:RAC2
120.57.102.103=DC1:RAC2
# Data Center Two
110.56.12.120=DC2:RAC1
110.50.13.201=DC2:RAC1
110.54.35.184=DC2:RAC1
50.33.23.120=DC2:RAC2
50.45.14.220=DC2:RAC2
50.17.10.203=DC2:RAC2
# Analytics Replication Group
172.106.12.120=DC3:RAC1
172.106.12.121=DC3:RAC1
172.106.12.122=DC3:RAC1
# default for unknown nodes
default=DC3:RAC1
Make sure the data center names defined in the /etc/dse/cassandra/cassandratopology.properties file correlates to what you name your data centers in your keyspace definition.
177
Deploying
Choosing keyspace replication options
When you create a keyspace, you must define the replica placement strategy class and the number of
replicas you want. DataStax recommends choosing NetworkTopologyStrategy for single and multiple data
center clusters. This strategy is as easy to use as the SimpleStrategy and allows for expansion to multiple
data centers in the future. It is much easier to configure the most flexible replication strategy up front, than
to reconfigure replication after you have already loaded data into your cluster.
NetworkTopologyStrategy takes as options the number of replicas you want per data center. Even for
single data center clusters, you can use this replica placement strategy and just define the number of
replicas for one data center. For example:
CREATE KEYSPACE test
WITH REPLICATION= { 'class' :
6 };
'NetworkTopologyStrategy',
'us-east' :
For a single node cluster, use the default data center name, Cassandra, Solr, or Analytics.
CREATE KEYSPACE test
WITH REPLICATION= { 'class' :
1 };
'NetworkTopologyStrategy',
'Analytics' :
To define the number of replicas for a multiple data center cluster:
CREATE KEYSPACE test2
WITH REPLICATION= { 'class' :
'dc2' : 3 };
'NetworkTopologyStrategy',
'dc1' : 3,
When creating the keyspace, what you name your data centers depends on the snitch you have chosen for
your cluster. The data center names must correlate to the snitch you are using in order for replicas to be
placed in the correct location.
As a general rule, the number of replicas should not exceed the number of nodes in a replication group.
However, it is possible to increase the number of replicas, and then add the desired number of nodes
afterwards. When the replication factor exceeds the number of nodes, writes will be rejected, but reads will
still be served as long as the desired consistency level can be met.
The default consistency level is QUORUM.
Changing replication settings
The default replication of 1 for keyspaces is suitable only for development and testing of a single node. For
production environments, it is important to change the replication of keyspaces from 1 to a higher number.
To avoid operations problems, changing the replication of these system keyspaces is especially important
for:
•
•
•
HiveMetaStore
cfs
cfs_archive
If the node is an Analytics node that uses Hive, increase the HiveMetaStore and cfs keyspace replication
factors to 2 or higher to be resilient to single-node failures. If you use cfs_archive, increase it accordingly.
Procedure
To change the replication keyspaces:
1. Check the name of the data center of the node:
•
•
Packaged installs: nodetool status
Tarball installs: install_location/bin/nodetool status
The output tells you the name of the data center for the node, for example, datacenter1.
2. Change the replication of the cfs and cfs_archive keyspaces from 1 to 3, for example:
178
Deploying
ALTER KEYSPACE cfs
WITH REPLICATION= {'class' : 'NetworkTopologyStrategy', 'dc1' : 3};
ALTER KEYSPACE cfs_archive
WITH REPLICATION= {'class' : 'NetworkTopologyStrategy', 'dc1' : 3};
How high you increase the replication depends on the number of nodes in the cluster, as discussed in
the previous section.
3. If you use Hive, update the HiveMetaStore keyspace to increase the replication from 1 to 3, for
example.
4. If the keyspaces you changed contain any data, run nodetool repair to avoid having missing data
problems or data unavailable exceptions.
Single data center deployment
A deployment scenario with a mixed workload cluster has only one data center for each type of workload.
About this task
In this scenario, a mixed workload cluster has only one data center for each type of node. For example, if
the cluster has 3 Hadoop nodes, 3 Cassandra nodes, and 2 Solr nodes, the cluster has 3 data centers, one
for each type of node. In contrast, a multiple data center cluster has more than one data center for each
type of node.
In single data center deployments, data replicates across the data centers automatically and transparently.
No ETL work is necessary to move data between different systems or servers. You can configure the
number of copies of the data in each data center and Cassandra handles the rest, replicating the data for
you.
To configure a multiple data center cluster, see Multiple data center deployment.
Before you begin
•
•
•
•
•
•
•
DataStax Enterprise is installed on each node.
Choose a name for the cluster.
For a mixed-workload cluster, determine the purpose of each node.
Get the IP address of each node.
Determine which nodes are seed nodes. (Seed nodes provide the means for all the nodes to find each
other and learn the topology of the ring.)
Other possible configuration settings are described in the cassandra.yaml configuration file.
Set virtual nodes correctly for the type of data center. DataStax recommends using virtual nodes only
on data centers running Cassandra real-time workloads. See Virtual nodes.
Procedure
This configuration example describes installing a 6 node cluster spanning 2 racks in a single data center.
The default consistency level is QUORUM.
1. Suppose the nodes have the following IPs and one node per rack will serve as a seed:
•
•
•
•
•
•
•
node0 110.82.155.0 (Cassandra seed)
node1 110.82.155.1 (Cassandra)
node2 110.82.155.2 (Cassandra)
node3 110.82.155.3 (Analytics seed)
node4 110.82.155.4 (Analytics)
node5 110.82.155.5 (Analytics)
node6 110.82.155.6 (Search - seed nodes are not required for Solr.)
179
Deploying
• node7 110.82.155.7 (Search)
2. If the nodes are behind a firewall, open the required ports for internal/external communication. See
Configuring firewall port access.
3. If DataStax Enterprise is running, stop the nodes and clear the data:
•
Packaged installs:
•
$ sudo service dse stop
$ sudo rm -rf /var/lib/cassandra/* ## Clears the data from the
directories
Tarball installs:
default
From the install directory:
$ sudo bin/dse cassandra-stop
$ sudo rm -rf /var/lib/cassandra/* ## Clears the data from the default
directories
Note: If you are clearing data from an AMI installation for restart, you need to preserve the
log files.
4. Set the properties in the cassandra.yaml file for each node.
Important: After making any changes in the cassandra.yaml file, you must restart the node
for the changes to take effect.
Location:
•
•
Packaged installs: /etc/dse/cassandra/cassandra.yaml
Tarball installs: install_location/resources/cassandra/conf/cassandra.yaml
Properties to set:
Note: If the nodes in the cluster are identical in terms of disk layout, shared libraries, and so on,
you can use the same copy of the cassandra.yaml file on all of them.
•
•
•
•
•
•
num_tokens: 256 for Cassandra nodes
num_tokens: 1 for Hadoop and Solr nodes
-seeds: internal_IP_address of each seed node
listen_address: empty
If not set, Cassandra asks the system for the local address, the one associated with its hostname.
In some cases Cassandra doesn't produce the correct address and you must specify the
listen_address.
auto_bootstrap: false (Add this setting only when initializing a fresh cluster with no data.)
If you are using a cassandra.yaml from a previous version, remove the following options, as they
are no longer supported by DataStax Enterprise:
## Replication strategy to use for the auth keyspace.
auth_replication_strategy: org.apache.cassandra.locator.SimpleStrategy
auth_replication_options:
replication_factor: 1
Example:
cluster_name: 'MyDemoCluster'
num_tokens: 256
seed_provider:
- class_name: org.apache.cassandra.locator.SimpleSeedProvider
parameters:
- seeds: "110.82.155.0,110.82.155.3"
listen_address:
5. After you have installed and configured DataStax Enterprise on all nodes, start the seed nodes one at a
time, and then start the rest of the nodes:
180
Deploying
•
•
Packaged installs: See Starting DataStax Enterprise as a service.
Tarball installs: See Starting DataStax Enterprise as a stand-alone process.
Note: If the node has restarted because of automatic restart, you must stop the node and clear
the data directories, as described above.
6. Check that your cluster is up and running:
•
•
Packaged installs: $ nodetool status
Tarball installs: $ install_location/bin/nodetool status
Results
Multiple data center deployment
A deployment scenario with a mixed workload cluster has more than one data center for each type of node.
About this task
In this scenario, a mixed workload cluster has more than one data center for each type of node. For
example, if the cluster has 4 Hadoop nodes, 4 Cassandra nodes, and 2 Solr nodes, the cluster could have
5 data centers: 2 data centers for Hadoop nodes, 2 data centers for Cassandra nodes, and 1 data center
for Solr nodes. A single data center cluster has only 1 data center for each type of node
In multiple data center deployments, data replication can be distributed across multiple, geographically
dispersed data centers; between different physical racks in a data center; or between public cloud
providers and on-premise managed data centers. Data replicates across the data centers automatically
and transparently. No ETL work is necessary to move data between different systems or servers. You can
configure the number of copies in each data center and Cassandra handles the rest, replicating the data
for you. For more information about replication:
•
•
Choosing keyspace replication options
Replication in a physical or virtual data center (Applies only to the single-token-per-node architecture.)
To configure a single data center cluster, see Single data center deployment.
Before you begin
To configure a multi-node cluster with multiple data centers:
•
•
•
•
•
•
•
•
DataStax Enterprise is installed on each node.
Choose a name for the cluster.
For a mixed-workload cluster, determine the purpose of each node.
Get the IP address of each node.
Determine which nodes are seed nodes. (Seed nodes provide the means for all the nodes to find each
other and learn the topology of the ring.)
Develop a naming convention for each data center and rack, for example: DC1, DC2 or 100, 200 and
RAC1, RAC2 or R101, R102.
Other possible configuration settings are described in the cassandra.yaml configuration file.
Set virtual nodes correctly for the type of data center. DataStax recommends using virtual nodes only
on data centers running Cassandra real-time workloads. See Virtual nodes.
181
Deploying
Procedure
This configuration example describes installing a 6 node cluster spanning 2 data centers. The default
consistency level is QUORUM.
1. Suppose you install DataStax Enterprise on these nodes:
• node0 10.168.66.41 (seed1)
• node1 10.176.43.66
• node2 10.168.247.41
• node3 10.176.170.59 (seed2)
• node4 10.169.61.170
• node5 10.169.30.138
2. If the nodes are behind a firewall, open the required ports for internal/external communication. See
Configuring firewall port access.
3. If DataStax Enterprise is running, stop the nodes and clear the data:
•
Packaged installs:
•
$ sudo service dse stop
$ sudo rm -rf /var/lib/cassandra/* ## Clears the data from the
directories
Tarball installs:
default
From the install directory:
$ sudo bin/dse cassandra-stop
$ sudo rm -rf /var/lib/cassandra/* ## Clears the data from the default
directories
Note: If you are clearing data from an AMI installation for restart, you need to preserve the
log files.
4. Set the properties in the cassandra.yaml file for each node.
Important: After making any changes in the cassandra.yaml file, you must restart the node
for the changes to take effect.
Location:
•
•
Packaged installs: /etc/dse/cassandra/cassandra.yaml
Tarball installs: install_location/resources/cassandra/conf/cassandra.yaml
Properties to set:
Note: If the nodes in the cluster are identical in terms of disk layout, shared libraries, and so on,
you can use the same copy of the cassandra.yaml file on all of them.
•
•
•
•
•
•
num_tokens: 256 for Cassandra nodes
num_tokens: 1 for Hadoop and Solr nodes
-seeds: internal_IP_address of each seed node
listen_address: empty
If not set, Cassandra asks the system for the local address, the one associated with its hostname.
In some cases Cassandra doesn't produce the correct address and you must specify the
listen_address.
auto_bootstrap: false (Add this setting only when initializing a fresh cluster with no data.)
If you are using a cassandra.yaml from a previous version, remove the following options, as they
are no longer supported by DataStax Enterprise:
## Replication strategy to use for the auth keyspace.
auth_replication_strategy: org.apache.cassandra.locator.SimpleStrategy
auth_replication_options:
182
Deploying
replication_factor: 1
Example:
You must include at least one seed node from each data center. It is a best practice to have more than
one seed node per data center.
cluster_name: 'MyDemoCluster'
num_tokens: 256
seed_provider:
- class_name: org.apache.cassandra.locator.SimpleSeedProvider
parameters:
- seeds: "10.168.66.41,10.176.170.59"
listen_address:
5. If necessary, change the dse.yaml file on each node to specify the snitch to be delegated by the
DseDelegateSnitch. For more information about snitches, see the About Snitches.
•
•
Packaged installs: /etc/dse/dse.yaml
Tarball installs: install_location/resources/dse/conf/dse.yaml
Example of specifying the PropertyFileSnitch:
delegated_snitch: org.apache.cassandra.locator.PropertyFileSnitch
6. In the cassandra-topology.properties file, use your naming convention to assign data center
and rack names to the IP addresses of each node, and assign a default data center name and rack
name for unknown nodes.
•
•
Packaged installs: /etc/dse/cassandra/cassandra-topology.properties
Tarball installs: install_location
Example:
# Cassandra Node IP=Data Center:Rack
10.168.66.41=DC1:RAC1
10.176.43.66=DC2:RAC1
10.168.247.41=DC1:RAC1
10.176.170.59=DC2:RAC1
10.169.61.170=DC1:RAC1
10.169.30.138=DC2:RAC1
# default for unknown nodes
default=DC1:RAC1
7. After you have installed and configured DataStax Enterprise on all nodes, start the seed nodes one at a
time, and then start the rest of the nodes:
•
•
Packaged installs: See Starting DataStax Enterprise as a service.
Tarball installs: See Starting DataStax Enterprise as a stand-alone process.
Note: If the node has restarted because of automatic restart, you must stop the node and clear
the data directories, as described above.
8. Check that your cluster is up and running:
•
•
Packaged installs: $ nodetool status
Tarball installs: $ install_location/bin/nodetool status
Results
183
Deploying
What to do next
•
•
Configuring system_auth keyspace replication
Replication in a physical or virtual data center (Applies only to the single-token-per-node architecture.)
Single-token architecture deployment
Use single-token architecture deployment when you are not using virtual nodes (vnodes).
About this task
Follow these steps only when you are not using virtual nodes (vnodes).
Before you begin
•
•
•
•
•
•
•
•
Have a basic understanding of tokens and Database internals.
DataStax Enterprise is installed on each node.
Choose a name for the cluster.
Take note of the total number of nodes in the cluster.
For a mixed-workload cluster, determine the purpose of each node.
Determine which nodes are seed nodes and record their IP addresses. Seed nodes provide the means
for all the nodes to find each other and learn the topology of the ring.
If using multiple data centers, develop a naming convention for each data center and rack, for example:
DC1, DC2 or 100, 200 and RAC1, RAC2 or R101, R102.
Other possible configuration settings are described in the cassandra.yaml configuration file.
Procedure
1. Suppose you install DataStax Enterprise on these nodes:
• node0 10.168.66.41 (seed1)
• node1 10.176.43.66
• node2 10.168.247.41
• node3 10.176.170.59 (seed2)
• node4 10.169.61.170
• node5 10.169.30.138
2. Calculate the token assignments using the information on Calculating tokens.
Table 2: Single Data Center
184
Node
Token
node0
0
node1
21267647932558653966460912964485513216
node2
42535295865117307932921825928971026432
node3
63802943797675961899382738893456539648
node4
85070591730234615865843651857942052864
node5
106338239662793269832304564822427566080
Deploying
Table 3: Multiple Data Centers
Node
Token
Offset
Data Center
node0
0
NA
DC1
node1
56713727820156410577229101238628035242
NA
DC1
node2
113427455640312821154458202477256070485
NA
DC1
node3
100
DC2
node4
56713727820156410577229101238628035342
100
DC2
node5
113427455640312821154458202477256070585
100
DC2
100
3. If the nodes are behind a firewall, open the required ports for internal/external communication. See
Configuring firewall port access.
4. If DataStax Enterprise is running, stop the nodes and clear the data:
•
Packaged installs:
•
$ sudo service dse stop
$ sudo rm -rf /var/lib/cassandra/* ## Clears the data from the
directories
Tarball installs:
default
From the install directory:
$ sudo bin/dse cassandra-stop
$ sudo rm -rf /var/lib/cassandra/* ## Clears the data from the default
directories
Note: If you are clearing data from an AMI installation for restart, you need to preserve the
log files.
5. Set the properties in the cassandra.yaml file for each node.
Important: After making any changes in the cassandra.yaml file, you must restart the node
for the changes to take effect.
Location:
•
•
Packaged installs: /etc/dse/cassandra/cassandra.yaml
Tarball installs: install_location/resources/cassandra/conf/cassandra.yaml
Properties to set:
•
•
•
•
initial_token: token
num_tokens: 1
-seeds: internal_IP_address of each seed node
listen_address: empty
•
If not set, Cassandra asks the system for the local address, the one associated with its hostname. In
some cases Cassandra doesn't produce the correct address and you must specify the list_address.
auto_bootstrap: false
•
Add the bootstrap setting only when initializing a fresh cluster with no data.
If you are using a cassandra.yaml from a previous version, remove the following options, as they
are no longer supported by DataStax Enterprise:
## Replication strategy to use for the auth keyspace.
auth_replication_strategy: org.apache.cassandra.locator.SimpleStrategy
auth_replication_options:
replication_factor: 1
185
Deploying
Example:
If using more than one data center, include at least one seed node from each data center. It is a best
practice to have more than one seed node per data center.
cluster_name: 'MyDemoCluster'
num_tokens: 256
seed_provider:
- class_name: org.apache.cassandra.locator.SimpleSeedProvider
parameters:
- seeds: "10.168.66.41,10.176.170.59"
listen_address:
6. If necessary, change the dse.yaml file on each node to specify the snitch to be delegated by the
DseDelegateSnitch. Be sure to use a data center and rack-aware snitch. See About Snitches.
•
•
Packaged installs: /etc/dse/dse.yaml
Tarball installs: install_location/resources/dse/conf/dse.yaml
Example of specifying the PropertyFileSnitch:
delegated_snitch: org.apache.cassandra.locator.PropertyFileSnitch
7. In the cassandra-topology.properties file, use your naming convention to assign data center
and rack names to the IP addresses of each node, and assign a default data center name and rack
name for unknown nodes.
•
•
Packaged installs: /etc/dse/cassandra/cassandra-topology.properties
Tarball installs: install_location
Example:
# Cassandra Node IP=Data Center:Rack
10.168.66.41=DC1:RAC1
10.176.43.66=DC2:RAC1
10.168.247.41=DC1:RAC1
10.176.170.59=DC2:RAC1
10.169.61.170=DC1:RAC1
10.169.30.138=DC2:RAC1
# default for unknown nodes
default=DC1:RAC1
8. After you have installed and configured DataStax Enterprise on all nodes, start the seed nodes one at a
time, and then start the rest of the nodes:
•
•
Packaged installs: See Starting DataStax Enterprise as a service.
Tarball installs: See Starting DataStax Enterprise as a stand-alone process.
Note: If the node has restarted because of automatic restart, you must stop the node and clear
the data directories, as described above.
9. Check that your cluster is up and running:
•
•
Packaged installs: $ nodetool status
Tarball installs: $ install_location/bin/nodetool status
Calculating tokens
Tokens assign a range of data to a particular node within a data center.
When you start a DataStax Enterprise cluster, you must choose how the data is divided across the nodes
in the cluster. A partitioner determines what each node stores by row. A token is a partitioner-dependent
element of the cluster. Each node in a cluster is assigned a token and that token determines the node's
position in the ring and what data the node is responsible for in the cluster. The tokens assigned to your
186
Deploying
nodes need to be distributed throughout the entire possible range of tokens. Each node is responsible for
the region of the ring between itself (inclusive) and its predecessor (exclusive). As a simple example, if the
range of possible tokens is 0 to 100 and there are 4 nodes, the tokens for the nodes would be: 0, 25, 50,
75. This approach ensures that each node is responsible for an equal range of data.
Note:
When using multiple data centers, each data center should be partitioned as if it were its own
distinct ring.
Before the node is started for the first time, each node in the cluster must be assigned a token. Set
the token with the initial_token property in the cassandra.yaml configuration file. Also, be sure to
comment out the num_token property.
For more detailed information, see Database internals.
Calculating tokens for a single data center
For example, for 6 nodes in a single data center, the results for calculating tokens using the
Murmur3Partitioner are:
[ '-9223372036854775808', '-6148914691236517206', '-3074457345618258604',
'-2', '3074457345618258600', '6148914691236517202' ]
Calculating tokens for multiple racks in a single data center
If you have multiple racks in single data center, calculate the tokens for the number of nodes and then
assign the tokens to nodes on alternating racks. For example: rack1, rack2, rack1, rack2, and so on. The
image shows the rack assignments:
As a best practice, each rack should have the same number of nodes so you can alternate the rack
assignments.
Calculating tokens for a multiple data center cluster
In multiple data center deployments using the NetworkTopologyStrategy, calculate the replica placement
for your custom keyspaces per data center. The NetworkTopologyStrategy determines replica placement
independently within each data center. The first replica is placed according to the partitioner. Additional
replicas in the same data center are determined by walking the ring clockwise until a node in a different
rack from the previous replica is found. If no such node exists, additional replicas are placed in the same
rack.
Do not use SimpleStrategy for this type of cluster. There are different methods you can use when
calculating multiple data center clusters. The important point is that the nodes within each data center
manage an equal amount of data; the distribution of the nodes within the cluster is not as important.
DataStax recommends using DataStax Enterprise OpsCenter to re-balance a cluster.
Alternating token assignments
187
Deploying
Calculate the tokens for each data center and then alternate the token assignments so that the nodes for
each data center are evenly dispersed around the ring. The following image shows the token position and
data center assignments:
Avoiding token collisions
To avoid token collisions, offset the values for each token. Although you can increment in values of 1, it is
better to use a larger offset value, such as 100, to allow room to replace a dead node.
The following shows the tokens for a cluster with two 3 node data centers and one 2 node data center.
Tokens for 3 nodes:
['-9223372036854775808',
'-3074457345618258603',
'3074457345618258602']
Tokens for 2 nodes:
['-9223372036854775808',
'0']
Using an offset value of 100:
•
Data Center 1
•
['-9223372036854775808',
'-3074457345618258603',
'3074457345618258602']
Data Center 2
•
['-9223372036854775708',
'-3074457345618258503',
'3074457345618258702']
Data Center 3
['-9223372036854775608',
'200']
Expanding a DataStax AMI cluster
To expand your EC2 implementations, use OpsCenter to provision a new cluster, add a new cluster, or add
nodes to a cluster.
The best way to expand your EC2 implementations is to use OpsCenter:
•
•
•
188
Provisioning a new cluster
Adding an existing cluster
Adding nodes to a cluster
DataStax Enterprise tools
DataStax Enterprise tools
Tools include dse commands, dsetool, dfs-stress tool, pre-flight check, yaml_diff, and the Cassandra bulk
loader.
The dse commands
Table of dse commands for using DataStax Enterprise
You can issue the dse commands listed in this document from the bin directory of the DataStax Enterprise
Linux installation or from the command line in a packaged or AMI distribution.
DSE commands
Synopsis
dse [-v ] | cassandra [options ] | hadoop [options ] | hive [options ]
| mahout [options ] | pig [options ] | sqoop [options ]
This table describes the key dse commands:
Command
Option
Description
Example
dse
-v
Send the DSE version number to
standard output.
none
Start up a real-time Cassandra node in
the background.
link to example
Start up a DSE Search/Solr node in the
background.
link to example
dse
cassandra
dse
cassandra
-s
dse
cassandra
-s Use path to store Solr data.
Ddse.solr.data.dir=path
link to example
dse
cassandra
-t
Start up an analytics node in the
background.
link to example
dse
cassandra
-t -j
Start up an analytics node as the job
tracker.
link to example
dse
cassandra
-f
Start up a real-time Cassandra node in
the foreground.
none
dse
cassandra
-f -t
Start up an analytics node in the
foreground.
none
dse
cassandra
-f -s
Start up a DSE Search/Solr node in the
foreground.
none
dse
cassandrastop
-p pid
Stop the DataStax Enterprise process
number pid.
link to example
dse
cassandra
After replacing a node, replace the IP
Dcassandra.replace_address
address in the table.
none
dse hadoop
version
none
Sends the version of the Hadoop
component to standard output.
189
DataStax Enterprise tools
Command
Option
Description
Example
dse hadoop
fs options
Invoke the Hadoop FileSystem shell.
link to example
dse hadoop
fs -help
Send Apache Hadoop fs command
descriptions to standard output.
link to example
Start the Hive client.
link to example
Start Hive by connecting through the
JDBC driver.
link to example
dse
options
list_subranges
List subranges of data in a keyspace.
link to documentation
dse mahout
Describe Mahout commands.
link to example
dse hive
dse hive
--service name
dse mahout
mahout command
options
Run the Mahout command.
link to example
dse mahout
hadoop
hadoop command
options
Add Mahout classes to classpath and
execute the hadoop command.
link to example
Start Pig.
link to example
Send Apache Sqoop command line help
to standard output.
link to example
dse pig
dse sqoop
-help
Hadoop, hive, mahout, and pig commands must be issued from an analytics node. The hadoop fs options,
which DSE Analytics supports with one exception (-moveToLocal), are described in the HDFS File System
Shell Guide on the Apache Hadoop web site. DSE Analytics has not yet implemented the -moveToLocal
option, but you can use the -copyToLocal.
The dsetool
Use the dsetool utility for Cassandra File System (CFS) and Hadoop-related tasks, such as managing the
job tracker, checking the CFS, and listing node subranges of data in a keyspace.
You can use the dsetool utility for CassandraFS- and Hadoop-related tasks, such as managing the job
tracker, checking the CassandraFS, and listing node subranges of data in a keyspace. In DataStax
Enterprise 4.0.2 and later, only JMX ( java management extensions) provides dsetool password
authentication. If JMX passwords are enabled, users then need to use the passwords to use the dsetool
utility.
Usage: dsetool [-h|--host=<hostname>] [-p|--port=<#>] [-j|--jmxport=<#>]
<command> <args>
This table describes the dsetool arguments:
Short form
Long form
Description
-a
--jmx-username <arg>
User name for authenticating with secure JMX
-b
--jmx-password <arg>
Password for authenticating with secure JMX
-h
--host <arg>
Node hostname or IP address
-j
--jmxport <arg>
Remote jmx agent port number
-u
--use_hadoop_config
Get cassandra host from hadoop configuration files
The dsetool commands are:
190
DataStax Enterprise tools
•
•
•
•
•
•
•
•
•
•
•
•
checkcfs - Check a single CassandraFS file or the whole CassandraFS.
createsystemkey <encryption option> [<encryption option ... >] [<system key name>] - Creates the
system key for transparent data encryption. DataStax 4.0.4 and later.
inmemorystatus <keyspace> <table> - Provides the memory size, capacity, and percentage used by the
table. The unit of measurement is MB. Bytes are truncated.
listjt - List all JobTracker nodes grouped by DC local to them.
list_subranges <keyspace> <cf-name> <keys_per_range> <start_token>, <end_token> - Divide a token
range for a given keyspace/table into a number of smaller subranges of approximately keys_per_range.
To be useful, the specified range should be contained by the target node's primary range.
jobtracker - Return the JobTracker hostname and port, JT local to the DC from which you are running
the command.
movejt - Move the JobTracker and notify the TaskTracker nodes.
partitioner - Return the fully qualified classname of the IPartitioner in use by the cluster
repaircfs - Repair the CFS from orphan blocks.
rebuild_indexes <keyspace> <table-name> <idx1,idx2,...> - Rebuild specified secondary indexes for
given keyspace/table. Use only keyspace/table-name to re-build all indexes.
ring - List the nodes in the ring including their node type.
status - Same as the ring command.
Examples of using dsetool commands for managing the Job Tracker are presented in Managing the job
tracker using dsetool commands.
Checking the CassandraFS using dsetool
Use the dsetool checkcfs command to scan the CassandraFS for corrupted files. For example:
dsetool checkcfs cfs:///
Use the dsetool to get details about a particular file that has been corrupted. For example:
dsetool checkcfs /tmp/myhadoop/mapred/system/jobtracker.info
Listing sub-ranges using dsetool
The dsetool command syntax for listing subranges of data in a keyspace is:
dsetool [-h ] [hostname ] list_subranges keyspace table rows per
subrange start token end token
•
•
•
rows per subrange is the approximate number of rows per subrange.
start partition range is the start range of the node.
end partition range is the end range of the node.
Note: You run nodetool repair on a single node using the output of list_subranges. The output
must be partition ranges used on that node.
Example
dsetool list_subranges Keyspace1 Standard1 10000
113427455640312821154458202477256070485 0
Output
The output lists the subranges to use as input to the nodetool repair command. For example:
Start Token
End Token
Estimated Size
----------------------------------------------------------------------------------------113427455640312821154458202477256070485
132425442795624521227151664615147681247 11264
132425442795624521227151664615147681247
151409576048389227347257997936583470460 11136
191
DataStax Enterprise tools
151409576048389227347257997936583470460 0
11264
Nodetool repair command options
You need to use the nodetool utility when working with sub-ranges. The start partition range (-st) and end
partition range (-et) options specify the portion of the node needing repair. You get values for the start and
end tokens from the output of dsetool list_subranges command. The new nodetool repair syntax for using
these options is:
nodetool repair keyspace table -st start token
-et end token
Example
nodetool repair Keyspace1 Standard1 -st
113427455640312821154458202477256070485 -et
132425442795624521227151664615147681247
nodetool repair Keyspace1 Standard1 -st
132425442795624521227151664615147681247 -et
151409576048389227347257997936583470460
nodetool repair Keyspace1 Standard1 -st
151409576048389227347257997936583470460 -et 0
These commands begins an anti-entropy node repair from the start partition range to the end partition
range.
Configuring the disk health checker
Ways to enable, disable, and use the disk health checker>
About this task
You can manage the disk health checker in DataStax Enterprise in the following ways:
•
•
•
Enable and disable the disk health checker.
Configure the frequency of disk health checking.
Manage the amount of disk space allowed on a partition for a data directory.
When the disk space exceeds the allowed amount, the node is shut down.
Configure the options in the dse.yaml file, which is located in:
•
•
Packaged installations: /etc/dse/dse.yaml
Tarball installations: install_location/resources/dse/conf/dse.yaml
health_check_interval option
By default, the disk health checker is enabled by DataStax Enterprise 4.0. The health_check_interval
option sets the time interval in minutes between each check. Any value greater than 0 enables the disk
health checker. By default, the health_check_interval is 5.
To disable the disk health checker, set the health_check_interval option to 0.
min_disk_space_before_failure
By default, the disk health checker allows 10 MB on a partition holding a data directory before shutting
down the node.
Pre-flight check and yaml_diff tools
The pre-flight check tool is available for packaged installations. This collection of tests can be run on a
node to detect and fix a configuration. The yaml_diff tool filters differences between cassandra.yaml files.
192
DataStax Enterprise tools
About this task
The pre-flight check tool, located in /usr/share/dse/tools of packaged installations, is a collection of
tests that can be run on a node to detect and fix a configuration. The tool can detect and fix many invalid or
suboptimal configuration settings. The tool is not available in tarball installations.
This release includes the yaml_diff tool that filters differences between two cassandra.yaml files, which is
useful during upgrades. The new tool is located in the tools directory.
Using the Cassandra bulk loader in a secure environment
Using sstableloader tool with Kerberos/SSL.
About this task
The Cassandra bulk loader is the sstableloader tool. The command-line options for configuring secure
sstableloader operations using Kerberos have changed slightly. If you run sstableloader from a DataStax
Enterprise node that has been configured for Kerberos or client-to-node/node-to-node encryption using
SSL, no additional configuration is needed for securing sstableloader operations. The sstableloader tool
will pick up all required options from the configured node automatically, so no further configuration is
needed. On an unconfigured developer machine, however, configure Kerberos or SSL as follows:
Kerberos
If you have not configured Kerberos on a DataStax Enterprise node, but you want to run sstableloader in a
secure Kerberos environment, set the options on the command line as follows:
•
•
To use credentials from default ticket cache, no extra options are necessary. sstableloader will do the
right thing.
To set the keytab location through system properties, use this example as a guide to setting the
options:
•
JVM_OPTS="-Dkerberos.use.keytab=true \
-Dkerberos.keytab=/home/dse/cassandra.keytab \
[email protected]" \
resources/cassandra/bin/sstableloader -d 192.168.56.102 /var/lib/
cassandra/data/Keyspace1/Standard1
To set Kerberos options using the JAAS config, use this example as a guide to setting the options:
•
JVM_OPTS="-Dkerberos.use.config.file=true \
-Djava.security.auth.login.config=/home/dse/keytab-basic-jaas.conf" \
resources/cassandra/bin/sstableloader -d 192.168.56.102 /var/lib/
cassandra/data/Keyspace1/Standard1
In the JAAS config, /home/dse/keytab-basic-jaas.conf, set these options:
Client {
com.sun.security.auth.module.Krb5LoginModule required
useKeyTab=true
keyTab="/home/dse/cassandra.keytab"
principal="[email protected]";
};
Client- and node-to-node encryption using SSL
If you have not configured SSL on a DataStax Enterprise node, but you want to run sstableloader in a
secure SSL environment, you can use the sstableloader script from Apache Cassandra to load SSTables
into a cluster with client-to-node/node-to-node SSL encryption enabled. Use the following basic options:
resources/cassandra/bin/sstableloader -d 192.168.56.102 /var/lib/cassandra/
data/Keyspace1/Standard1 \
-tf org.apache.cassandra.thrift.SSLTransportFactory \
193
DataStax Enterprise tools
-ts /path/to/truststore \
-tspw truststore_password
If you want to configure require_client_auth=true on the target, set these additional options:
resources/cassandra/bin/sstableloader -d 192.168.56.102 /var/lib/cassandra/
data/Keyspace1/Standard1 \
-tf org.apache.cassandra.thrift.SSLTransportFactory \
-ts /path/to/truststore \
-tspw truststore_password \
-ks /path/to/keystore \
-kspw keystore_password
194
Moving data to or from other databases
Moving data to or from other databases
DataStax offers several solutions for migrating from other databases.
About this task
DataStax offers several solutions for migrating from other databases:
•
•
•
•
The COPY command, which mirrors what the PostgreSQL RDBMS uses for file/export import.
Apache Sqoop, which is a tool designed to transfer data between an RDBMS and Hadoop. DataStax
Enterprise modified Sqoop so you can not only transfer data from an RDBMS to a Hadoop node in a
DataStax Enterprise cluster, but also move data directly into Cassandra as well.
The DSE Search/Solr Data Import Handler, which is a configuration-driven method for importing data to
be indexed for searching.
The Cassandra bulk loader provides the ability to bulk load external data into a cluster.
About the COPY command
You can use COPY in Cassandra’s CQL shell to load flat file data into Cassandra (nearly all RDBMS’s
have unload utilities that allow table data to be written to OS files) as well as data to be written out to OS
files.
ETL Tools
If you need more sophistication applied to a data movement situation (more than just extract-load), then
you can use any number of extract-transform-load (ETL) solutions that now support Cassandra. These
tools provide excellent transformation routines that allow you to manipulate source data in literally any way
you need and then load it into a Cassandra target. They also supply many other features such as visual,
point-and-click interfaces, scheduling engines, and more.
Many ETL vendors who support Cassandra supply community editions of their products that are free
and able to solve many different use cases. Enterprise editions are also available that supply many other
compelling features that serious enterprise data users need.
You can freely download and try ETL tools from Jaspersoft, Pentaho, and Talend that all work with
DataStax Enterprise and community Cassandra.
195
Reference
Reference
Reference topics.
Installing glibc on Oracle Linux 6.x and later
To install DSE on Oracle Enterprise Linux 6.x and later, install the 32-bit versions of the glibc libraries.
About this task
To install DataStax Enterprise on Oracle Enterprise Linux 6.x and later, you need to install the 32-bit
versions of the glibc libraries. They are not installed by default.
Procedure
1. Make the yum.repos.d your current directory.
$ cd /etc/yum.repos.d
2. Download the public-yum-ol6.repo package from the repository.
$ wget http://public-yum.oracle.com/public-yum-ol6.repo
3. Check that glibc.i686 is ready for installation and install it.
$ yum list
$ yum install glibc.i686
File locations for binary tarball installs
Locations when DataStax Enterprise was installed from a tarball.
For package installs, see File locations.
Directories for cassanda.yaml and dse.yaml
•
•
install_location/resources/cassandra/conf/cassandra.yaml
install_location/resources/dse/conf/dse.yaml
DataStax Enterprise directories
The tarball installation creates the following directories in the install_location directory:
Directories
Description
bin
DataStax Enterprise start scripts
datastax-agent
DataStax agent files
demos
DataStax demos
interface
javadoc
lib
196
resources/cassandra/bin
Cassandra utilities
resources/cassandra/conf
Cassandra configuration files
Reference
Directories
Description
resources/hadoop
Hadoop installation
resources/hive
Hive installation
resources/log4j-appender
log4j logging
resources/mahout
Mahout installation
resources/pig
Pig installation
resources/solr
Solr installation
resources/sqoop
Sqoop installation
resources/cassandra
Notices
resources/solr/web/solr/WEB-INF
SPENGO configuration
File locations for package installs
Locations when DataStax Enterprise was installed from a package.
For tarball installs, see File locations.
Directories for cassanda.yaml and dse.yaml
•
•
/etc/dse/cassandra/cassandra.yaml
/etc/dse/dse.yaml
Cassandra directories
Directories
Description
/var/lib/cassandra
Cassandra and CassandraFS data directories
/var/log/cassandra
/var/run/cassandra
/usr/share/dse/cassandra
Cassandra environment settings
/usr/share/dse/cassandra/lib
/usr/share/dse-demos
Portfolio, Solr, Sqoop demos
/usr/bin
usr/sbin
/etc/dse/cassandra/
Cassandra configuration files
/etc/init.d
/etc/security/limits.d
/etc/default/
/usr/share/doc/dse-libcassandra
Notices and cqlshrc samples
197
Reference
Hadoop directories
Directories
Description
/usr/share/dse/hadoop
Hadoop environment settings
/etc/dse/hadoop
Hadoop configuration files
Hive directories
Directories
Description
/usr/share/dse/hive
Hive environment settings
/etc/dse/hive
Hive configuration files
Mahout directories
Directories
Description
/usr/share/dse/mahout
Mahout properties files
/etc/dse/mahout
Mahout JAR files
Pig directories
Directories
Description
/usr/share/dse/pig
Pig environment settings
/etc/dse/pig
Pig configuration files
Solr directories
Directories
Description
/usr/share/dse-demos
Search - Wikipedia demo
/usr/share/dse/solr/web/solr/WEB-INF
SPENGO configuration
Sqoop directories
Directories
Description
/usr/share/dse/sqoop
Sqoop environment settings
/etc/dse/sqoop
/usr/share/dse-demos
Sqoop demo
Log4j directories
198
Directories
Description
/etc/dse/log4j
log4j configuration file
/usr/share/dse-demos
Log Search demo
Reference
OpsCenter directories
Directories
Description
/var/lib/opscenter
SSL certificates for encrypted agent/dashboard
communications
/var/log/opscenter
Log directory
/var/run/opscenter
Runtime files
/usr/share/opscenter
JAR, agent, web application, and binary files
/etc/opscenter
Configuration files
/etc/init.d
Service startup script
/etc/security/limits.d
OpsCenter user limits
DataStax Agent directories
Directories
Description
/var/lib/datastax-agent/ssl
SSL certificates for encrypted agent and dashboard
communications
/var/lib/datastax-agent/conf
Configuration file
/var/log/datastax-agent
Log directory
/var/run/datastax-agent
Runtime files
/usr/share/datastax-agent
JAR, agent, web application, and binary files
/etc/init.d
Service startup script
DataStax Enterprise configuration file (dse.yaml)
The configuration file for Kerberos authentication, purging of expired data from the Solr indexes, and
setting Solr inter-node communication.
The dse.yaml file is the configuration file for setting the delegated snitch, Kerberos authentication, and
purging of expired data from the Solr indexes. It is located in the following directories:
•
•
Packaged installs: /etc/dse
Tarball installs: install_location/resources/dse/conf
For cassandra.yaml configuration, see Node and cluster configuration (cassandra.yaml).
Snitch settings
The delegated_snitch property sets which snitch is delegated. For example, it sets the DseSimpleSnitch.
•
delegated_snitch
•
(Default: com.datastax.bdp.snitch.DseSimpleSnitch) - Sets which snitch is used.
DseSimpleSnitch
The DseSimpleSnitch places Cassandra, Hadoop, and Solr nodes into separate data centers. See
DseSimpleSnitch.
For more information, see Snitches in the Cassandra documentation.
199
Reference
Kerberos support
The kerberos_options set the QOP (Quality of Protection) and encryption options.
Options:
•
•
•
•
keytab: resources/dse/conf/dse.keytab
service_principal: dse/[email protected]
http_principal: HTTP/[email protected]
qop - auth A comma-delimited list of Quality of Protection values that clients and servers can use for
each connection. The valid values are:
•
•
•
auth - (Default) Authentication only.
auth-int - Authentication plus integrity protection for all transmitted data.
auth-conf - Authentication plus integrity protection and encryption of all transmitted data.
Encryption using auth-conf is separate and completely independent of whether encryption is done
using SSL. If both auth-conf and SSL are enabled, the transmitted data is encrypted twice. DataStax
recommends choosing one and using it for both encryption and authentication.
Multi-threaded indexing
DSE Search provides multi-threaded indexing implementation to improve performance on multi-core
machines. All index updates are internally dispatched to a per-core indexing thread pool and executed
asynchronously, which allows for greater concurrency and parallelism. However, index requests can return
a response before the indexing operation is executed.
max_solr_concurrency_per_core
(Default: number of available Solr cores times 2) Configures the maximum number of concurrent
asynchronous indexing threads per Solr core. If set to 1, DSE Search returns to the synchronous indexing
behavior.
back_pressure_threshold_per_core
(Default: 10000) The total number of queued asynchronous indexing requests per Solr core, computed at
Solr commit time. When exceeded, back pressure prevents excessive resources consumption by throttling
new incoming requests.
Scheduler settings for Solr indexes
These settings control the schedulers in charge of querying for and removing expired data.
ttl_index_rebuild_options
Options:
•
•
•
fix_rate_period - (Default: 300 seconds) Schedules how often to check for expired data.
initial_delay - (Default: 20 seconds) Speeds up startup by delaying the first TTL checks.
max_docs_per_batch - (Default: 200) The maximum number of documents deleted per batch by the
TTL rebuild thread.
flush_max_time_per_core
(Default: 5 minutes) The maximum time to wait before flushing asynchronous index updates, which occurs
at either at Solr commit time or at Cassandra flush time. To fully synchronize Solr indexes with Cassandra
data, ensure that flushing completes successfully by setting this value to a reasonable high value.
Starting and stopping DataStax Enterprise
Starting and stopping DataStax Enterprise as a service or stand-alone process.
200
Reference
After you have installed and configured DSE on one or more nodes, you are ready to start your cluster
starting with the seed nodes. In a mixed-workload DSE cluster, you must start the analytics seed node first.
Packaged installations include startup and stop scripts for running DSE as a service. Binary packages do
not.
Starting DataStax Enterprise as a service
Starting the DataStax Enterprise service when DataStax Enterprise was installed from a package.
About this task
Packaged installations provide startup scripts in /etc/init.d for starting DSE as a service.
For mixed-workload clusters for Cassandra-only nodes, skip step 1.
Procedure
1. Edit the /etc/default/dse file, and then edit the appropriate line to this file, depending on the type
of node you want:
•
•
HADOOP_ENABLED=1 - Designates the node as DataStax Enterprise analytics and starts the Hadoop
Job Tracker and Task Tracker services.
SOLR_ENABLED=1 - Starts the node as DSE Enterprise Search/Solr node.
Note: DataStax does not support using the SOLR_ENABLED and HADOOP_ENABLED
options to designate the same node for both search and Hadoop analytics.
2. Start the DataStax Enterprise service and agent:
$ sudo service dse start
$ sudo service datastax-agent start
3. To check if your cluster is up and running:
$ nodetool status
On RHEL and CentOS, the DSE service runs as a Java process. On Debian and Ubuntu systems, the
DSE service runs as a jsvc process.
What to do next
Verify DataStax Enterprise is running
Starting DataStax Enterprise as a stand-alone process
Starting the DataStax Enterprise process when DataStax Enterprise was installed from a tarball.
About this task
If running a mixed-workload cluster, determine which nodes to start as analytics, Cassandra, and search
nodes. Begin with the seed nodes first — analytics seed node, followed by the Cassandra seed node —
then start the remaining nodes in the cluster one at a time. For additional information, see Multiple data
center deployment.
Attention: Do not start all the nodes at once, because this causes contention among nodes to
become the jobtracker.
Procedure
1. From the install directory:
•
•
•
Real-time Cassandra node: $ bin/dse cassandra
Analytics node: $ bin/dse cassandra -t
Solr node: $ bin/dse cassandra -s
201
Reference
Note: DataStax does not support using the -t search tracker option in combination with the s option to mark the node for Hadoop analytics and search.
2. Start the DataStax agent:
$ ./datastax-agent/bin/datastax-agent
3. To check that your ring is up and running:
$ cd install_location
$ bin/nodetool status
If you are running an analytics node, there are several methods for designating the job tracker node.
What to do next
Verify DataStax Enterprise is running
Stopping a DataStax Enterprise node
Stopping DataStax Enterprise and the DataStax Agent on a node.
About this task
To speed up the restart process, before stopping the dse service, run nodetool drain. This step writes
the current memtables to disk. When you restart the node, Cassandra does not need to read through the
commit log. If you have durable writes set to false, which is unlikely, there is no commit log and you must
drain the node to prevent losing data.
Running nodetool drain before using the cassandra-stop command to stop a stand-alone process is
pointless because the cassandra-stop command drains the node before stopping it.
To stop the service on a node:
$ nodetool drain -h host name
$ sudo service dse stop
To stop the stand-alone process on a node:
From the install location:
$ install_location/bin/dse cassandra-stop ## Use sudo if necessary
In the unlikely event that the cassandra-stop command fails because it cannot find the process DataStax
Enterprise Java process ID (PID), the output instructs you to find the DataStax Enterprise Java process ID
(PID) manually, and stop the process using its PID number.
$ ps auwx | grep dse
$ bin/dse cassandra-stop -p PID
Verify DataStax Enterprise is running
Use nodetool status to verify that DataStax Enterprise is running.
About this task
Use nodetool status as it is much less verbose than nodetool ring.
•
•
Packaged installs: $ nodetool status
Binary installs: $ install_location/bin/nodetool status
Troubleshooting
Troubleshooting examples are useful to discover and resolve problems with DSE. Also check the
Cassandra troubleshooting documentation.
202
Reference
The following common problems, solutions, or workarounds have been reported about using DataStax
Enterprise. Be sure to also check the Cassandra troubleshooting documentation.
Mahout Jobs that Use Lucene Not Supported
DataStax does not currently support Mahout jobs, such as built-in support for creating vectors from Lucene
indexes, that use Lucene features. Attempting to run Mahout jobs that use Lucene features results in this
type of error message:
Error: class org.apache.mahout.vectorizer.
DefaultAnalyzer overrides final method
tokenStream.
MX4J warning message during installation
When Cassandra loads, you may notice a message that MX4J will not load and that mx4j-tools.jar is not in
the classpath.
You can ignore this message. MX4j provides an HTML and HTTP interface to JMX and is not necessary to
run Cassandra. DataStax recommends using OpsCenter. It has more monitoring capabilities than MX4J.
DSE Search/Solr cannot find custom files
Open Source Solr (OSS) supports relative paths set by the <lib> property in the solrconfig.xml, but
DSE Search/Solr does not. Configuring the Solr library path describes a workaround for this issue.
Cassandra Log4j appender solutions
DataStax Enterprise allows you to stream your web and application log information into a database cluster
via Apache log4j.
About this task
DataStax Enterprise allows you to stream your web and application log information into a database cluster
via Apache log4j.
About the log4j Utility
Apache log4j is a Java-based logging framework that provides runtime application feedback. It
provides the ability to control the granularity of log statements using an external configuration file
(log4j.properties).
With the Cassandra Appender, you can store the log4j messages in a table where they're available for indepth analysis using the Hadoop and Solr capabilities provided by DataStax Enterprise. For information
about Cassandra logging, see Logging Configuration. Additionally, DataStax provides a Log4j Search
Demo.
The log4j utility has three main components: loggers, appenders, and layouts. Loggers are logical log file
names. They are the names known to the Java application. Each logger is independently configurable for
the level of logging. Outputs are controlled by Appenders. Numerous Appenders are available and multiple
Appenders can be attached to any Logger. This makes it possible to log the same information to multiple
outputs. Appenders use Layouts to format log entries. In the example below, messages show the level, the
thread name, the message timestamp, the source code file, the line number, and the log message.
Log levels
The available levels are:
•
•
All - turn on all logging
OFF - no logging
203
Reference
•
•
•
•
•
•
FATAL - severe errors causing premature termination
ERROR - other runtime errors or unexpected conditions
WARN - use of deprecated APIs, poor use of API, near errors, and other undesirable or unexpected
runtime situations
DEBUG - detailed information on the flow through the system
TRACE - more detailed than DEBUG
INFO - highlight the progress of the application at a coarse-grained level
Datastax does not recommend using TRACE or DEBUG in production due to verbosity and performance.
Log messages
As mentioned above, the messages that appear in the log are controlled via the conf/log4j.properties
file. Using this properties file, you can control the granularity to the Java package and class levels. For
example, DEBUG messages from a particular class can be included in the log while messages from
others remain at a higher level. This is helpful to reduce clutter and to identify messages. The log is most
commonly a file and/or stdout. The format, behavior (such as file rolling), and so on is also configurable at
runtime.
Below are sample log messages from a Cassandra node startup:
INFO [main ] 2012-02-10 09:15:33,112 DatabaseDescriptor.java (line 495 )
Found table data in data directories. Consider using the CLI to define
your schema.
INFO [main ] 2012-02-10 09:15:33,135 CommitLog.java (line 166 )
No commitlog files found; skipping replay
INFO [main ] 2012-02-10 09:15:33,150 StorageService.java (line 400 )
Cassandra version: 1.0.7
INFO [main ] 2012-02-10 09:15:33,150 StorageService.java (line 401 )
Thrift API version: 19.20.0
INFO [main ] 2012-02-10 09:15:33,150 StorageService.java (line 414 )
Loading persisted ring state
...
Storing log4j messages in a table
The Cassandra Appender provides the capability to store log4j messages in a Cassandra table.
Procedure
1. Add resources/log4j-appender/lib/ to your application classpath.
2. Modify the conf/log4j.properties file, as shown in the example below:
# Cassandra Appender
log4j.appender.CASS = com.datastax.logging.appender.CassandraAppender
log4j.appender.CASS.hosts = 127.0.0.1
log4j.appender.CASS.port = 9160
#log4j.appender.CASS.appName = "myApp"
#log4j.appender.CASS.keyspaceName = "Logging"
#log4j.appender.CASS.columnFamily = "log_entries"
#log4j.appender.CASS.placementStrategy =
"org.apache.cassandra.locator.NetworkTopologyStrategy"
#log4j.appender.CASS.strategyOptions = {"DC1" : "1", "DC2" : "3" }
#log4j.appender.CASS.replicationFactor = 1
#log4j.appender.CASS.consistencyLevelWrite = ONE
#log4j.appender.CASS.maxBufferedRows = 256
log4j.logger.com.foo.bar = INFO, CASS
Commented lines are included for reference and to show the default values.
•
204
log4j.appender.CASS=com.datastax.logging.appender.CassandraAppender specifies
the CassandraAppender class and assigns it the CASS alias. This alias is referenced in the last line.
Reference
•
•
•
log4j.appender.CASS.hosts = 127.0.0.1 allows using a comma delimited list of Cassandra
nodes (in case a node goes down).
Specify replication options in:
log4j.appender.CASS.placementStrategy =
"org.apache.cassandra.locator.NetworkTopologyStrategy"
log4j.appender.CASS.strategyOptions = {"DC1" : "1", "DC2" : "3" }.
log4j.logger.com.foo.bar = INFO, CASS specifies that all log messages of level INFO and
higher, which are generated from the classes and sub-packages within the com.foo.barpackage,
are sent to the Cassandra server by the Appender.
By default, the CassandraAppender records log messages in the table log_entries in the Logging
keyspace. The definition of this table is as follows:
cqlsh:Logging> DESCRIBE TABLE log_entries;
CREATE TABLE log_entries (
KEY uuid PRIMARY KEY,
app_start_time bigint,
app_name text,
class_name text,
file_name text,
level text,
line_number text,
log_timestamp bigint,
logger_class_name text,
host_ip text,
host_name text,
message text,
method_name text,
ndc text,
thread_name text,
throwable_str_rep text
) WITH
comment = '' AND
comparator = text AND
row_cache_provider = 'ConcurrentLinkedHashCacheProvider' AND
key_cache_size = 200000.000000 AND
row_cache_size = 0.000000 AND
read_repair_chance = 1.000000 AND
gc_grace_seconds = 864000 AND
default_validation = text AND
min_compaction_threshold = 4 AND
max_compaction_threshold = 32 AND
row_cache_save_period_in_seconds = 0 AND
key_cache_save_period_in_seconds = 14400 AND
replication_on_write = True;
Example
Consider the following log snippet:
09:20:55,470 WARN SchemaTest:68 - This is warn message #163
09:20:55,470 INFO SchemaTest:71 - This is info message #489
09:20:55,471 ERROR SchemaTest:59 - Test exception.
java.io.IOException: Danger Will Robinson, Danger!
at com.datastax.logging.SchemaTest.testSavedEntries
(SchemaTest.java:58 )
at sun.reflect.NativeMethodAccessorImpl.invoke0 (Native Method )
...
Note that the ERROR entry above includes the stack trace associated with an Exception. The
associated rows in the log_entries table appear as follows (queried using cqlsh):
205
Reference
KEY,eea1256e-db24-4cef-800b-843b3b2fb72c | app_start_time,1328894454774 |
level,WARN |
log_timestamp,1328894455391 | logger_class_name,org.apache.log4j.Category |
message,
This is warn message #163 | thread_name,main |
KEY,f7283a71-32a2-43cf-888a-0c1d3328548d | app_start_time,1328894454774 |
level,INFO |
log_timestamp,1328894455064 | logger_class_name,org.apache.log4j.Category |
message,
This is info message #489 | thread_name,main |
KEY,37ba6b9c-9fd5-4dba-8fbc-51c1696bd235 | app_start_time,1328894454774 |
level,ERROR |
log_timestamp,1328894455392 | logger_class_name,org.apache.log4j.Category |
message,
Test exception. | thread_name,main | throwable_str_rep,java.io.IOException:
Danger
Will Robinson, Danger!
at com.datastax.logging.SchemaTest.testSavedEntries
(SchemaTest.java:58 )
at sun.reflect.NativeMethodAccessorImpl.invoke0 (Native Method )
Not all columns have values because the set of values in logging events depends on the manner
in which the event was generated, that is, which logging method was used in the code and the
configuration of the table.
Storing logging information in Cassandra provides the capability to do in-depth analysis via the
DataStax Enterprise platform using Hadoop and Solr.
Log4j search demo
The Log4j Search Demo shows an example of searching and filtering log4j messages generated by a
standard Java application.
About this task
The Log4j Search Demo shows an example of searching and filtering log4j messages generated by a
standard Java application. In the demo, a Hadoop pi calculation is run with a log4j.properties file set to use
the CassandraAppender included in DataStax Enterprise. As the logs are generated, they are indexed in
real time by Solr and made available for searching in the demo user interface.
For information on configuring log4j, see Cassandra Log4j appender solutions.
Before starting this demo, be sure that you have started DSE Search/Solr on a single node.
Running the demo
Procedure
1. Open a shell window or tab and make the log_search directory your current directory. The location of
the demo directory depends on your platform:
RHEL or Debian installations:
$ cd /usr/share/dse-demos/log_search
Tar distribution, such as Mac:
$ cd $DSE_HOME/demos/log_search
2. Open another shell window or tab and add the schema:
$ ./1-add-schema.sh
206
Reference
The script posts solrconfig.xml and schema.xml to these locations:
http://localhost:8983/solr/resource/Logging.log_entries/solrconfig.xml
http://localhost:8983/solr/resource/Logging.log_entries/schema.xml
3. Start a Hadoop job using demo's log4j settings:
$ ./2-run-hadoop-test.sh
4. Open the results in a web browser, where you can view and search for messages:
http://localhost:8983/demos/log_search/
5. Use the search (filter) feature to view the log messages.
207
Release notes
Release notes
Release notes for DataStax Enterprise 4.0.x.
Release notes cover these releases:
•
•
•
•
•
•
•
DataStax Enterprise 4.0.6
DataStax Enterprise 4.0.5
DataStax Enterprise 4.0.4
DataStax Enterprise 4.0.3
DataStax Enterprise 4.0.2
DataStax Enterprise 4.0.1
DataStax Enterprise 4.0
DataStax Enterprise 4.0.6 release notes
Components
•
Cassandra 2.0.11.93
Resolved issues
•
JMX defaults to binding only localhost to mitigate "CVE-2015-0225: Apache Cassandra remote
execution of arbitrary code." After you upgrade to DataStax Enterprise 4.0.6, get the latest default
values from cassandra-env.sh and ensure that your local cassandra-env.sh file has the latest
default values. If you require remote access of JMX, you can safely enable remote JMX access by
turning on JMX security. The latest default values from cassandra-env.sh are:
LOCAL_JMX=yes
if [ "$LOCAL_JMX" = "yes" ]; then
JVM_OPTS="$JVM_OPTS -Dcassandra.jmx.local.port=$JMX_PORT -XX:
+DisableExplicitGC"
else
JVM_OPTS="$JVM_OPTS -Dcom.sun.management.jmxremote.port=$JMX_PORT"
JVM_OPTS="$JVM_OPTS -Dcom.sun.management.jmxremote.rmi.port=$JMX_PORT"
JVM_OPTS="$JVM_OPTS -Dcom.sun.management.jmxremote.ssl=false"
JVM_OPTS="$JVM_OPTS -Dcom.sun.management.jmxremote.authenticate=true"
JVM_OPTS="$JVM_OPTS -Dcom.sun.management.jmxremote.password.file=/etc/
cassandra/jmxremote.password"
fi
DataStax Enterprise 4.0.5 release notes
DataStax Enterprise 4.0.5 includes updated components, enhancements and changes, and resolved
issues.
Updated components
Cassandra has been updated to 2.0.11.92.
Enhancements and changes
•
•
•
•
The default Kerberos authenticator now supports the Digest authentication.
If you use transparent data encryption, you need to run Opscenter agent under the same account as
the DataStax Enterprise user.
DSE Search now logs a warning when users do not map dynamic fields correctly.
DataStax Enterprise 4.0.5 includes for experts only the capability to tune index size relative to query
range speed by adjusting the precision step of DataStax internal field types.
Resolved issues
208
Release notes
•
•
•
•
•
•
•
Resolved the issue causing entries entries to show up in system.peers for decommissioned nodes.
(DSP-3968)
Fixed a problem causing an out-of-memory condition when running nodetool repair. (DSP-4104,
CASSANDRA-7983)
The GossipFilePropertySnitch and EC2MultiRegionSnitch now use private IP addressses
for communication during node repair between nodes in the same data center. (DSP-4192,
CASSANDRA-8084)
Resolved the CQL TRUNCATE command problem related to memory-only tables. (DSP-4135)
Querying partitions directly now works. (DSP-4293)
Fixed an issue causing a null pointer exception on non Solr workload nodes holding Solr data and
attempting to run the nodetool cleanup command on data. (DSP-4310)
Fixed a bug causing core creation to fail if the Cassandra timeuuid type is used inside a list, set, or map
collection. (DSP-4288)
DataStax Enterprise 4.0.4 release notes
DataStax Enterprise 4.0.4 includes updates to components, enhancements and changes, resolved issues,
patches, and issues. The following Hive issue (DSP-3534), resolved in this release, might break existing
code:
Possible code breaker
Hive will now correctly return NULL in places it was previously returning an empty string.
If your application inserts null string-type values into a table, adjust the code to expect NULL instead of an
empty string.
Updated Components
•
•
•
From Apache Cassandra 2.0.7 to Apache Cassandra 2.0.9.61
From Apache Hadoop 1.0.4.10 to Apache Hadoop 1.0.4.13
From Cassandra Java Driver 2.0.1 to a patched Cassandra Java Driver 2.0.4 (DSP-3848)
Enhancements and changes
•
General enhancements
•
•
The procedure for encrypting data has changed in this release. DataStax recommends migrating
data encrypted in earlier releases to DataStax Enterprise 4.0.4.
• This release includes an example of using Cassandra triggers in the demos directory.
• The _HOST macro has been changed to force the hostname part of the Kerberos service principal
to lowercase. The Kerberos service principal is generated from the host name, which can cause a
problem connecting to cqlsh running if lowercase host names are not used.
Hive enhancements
•
•
Support for the native protocol in Hive including the addition of 19 new Hive TBLPROPERTIES to
support the native protocol.
• A new cql3.partition.key property that maps Hive tables to CQL compound primary keys and
composite partition keys, as shown in these examples: Work with an unsupported data type and Use
a composite partition key
Pig changes
•
•
Cql Paging Recorder Reader has been removed from Cassandra. The CqlStorage handler is no
longer compatible with compact tables with clustering columns. Users who are accessing these
tables need to migrate the tables to CqlNativeStorage format. This format uses the Cql Record
Reader.
The CqlStorage handler is deprecated and slated for removal at some point in the future. Use the
CqlNativeStorage handler and cql:// for new pig applications.
Resolved issues
209
Release notes
DataStax Enterprise 4.0.4 fixes the following issues:
•
•
•
•
The DSE startup script overwrote the JAVA_LIBRARY_PATH environment variable instead of
appending it. (DSP-2914)
Solr blob data types used in composite keys threw an invalidating error. (3033)
The heap was not dumped on an out-of-memory condition. (DSP-3308)
Tomcat blocked shutdown when an out of memory error occurred on Solr nodes. (DSP-3328)
No value was returned when the fl request parameter specified a dynamic field in Thrift and the 'fl'
parameter had only dynamic fields. (DSP-3332)
Alteration of MapReduce output returned multiple results (MAPREDUCE-1597), which caused large
gzip files with a custom input format to be read multiple times. (DSP-3384)
Solr HTTP requests ignored permissions set on keyspaces and tables. (DSP-3388)
Solr query POST parameters were not logged in the DSE audit log. (DSP-3440)
Audit logging threw a null pointer exception if a prepared statement used a null value. (DSP-3447)
Hive returned NULL in places it previously returned an empty string. See Possible code breaker.
(DSP-3534)
Creation of a Solr cql core using the CLUSTER ORDERING directive and a TimeUUID threw an
IllegalStateException. (DSP-3542)
Soft commit time interval reverted to a previous setting if changed during back pressure. (DSP-3584)
Reloading during back pressure of the old pre-reload value of the soft commit. (DSP-3584)
Removal of a node from the ring by the disk full alert feature.
•
DataStax Enterprise 3.x added a feature to auto-decommission a node when the disk is close to being
full. This feature has been removed and there are no plans to do any sort of proactive cluster changes
in the future. Because of the adverse effects from filling up data disks, customers are always advised to
monitor disk space and add capacity when necessary. (DSP-3601)
An out-of-memory error when using the Cassandra File System. (CFS)
•
Memory consumption has been reduced significantly, by a factor of approximately 500 for some use
cases. (DSP-3615)
Lack of support for writing to blob columns from Spark.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
This release supports reading columns of all types; however, you need to convert collections of blobs to
byte arrays before serializing. (DSP-3620)
The TermVectorComponent, not fixed in DSP-3147, caused unnecessary Cassandra access in
CassandraRowReader due to the StoredFieldVisitor returning an empty collection of fields to load.
(DSP-3625)
Failed repair, addition, and removal of nodes from a cluster using transparent data encryption.
When attempting to repair a node, the node did not unencrypt the data for reading. (DSP-3636).
Improper loading of SSTables with encrypted tables using sstableloader when internal authentication,
SSL, or kerberos authentication. (DSP-3641)
DataStax Enterprise failed to find the org.tartarus.snowball.ext.DanishStemmer class, which prevented
clustering in Solr. (DSP-3645)
A race condition between shutdown of client and server channels caused a harmless Netty exception to
appear. (DSP-3651)
Hive was incapable of reading from a varint column if Cassandra was configured to use the random
partitioner (DSP-3652)
A null pointer exception occurred when using QueryBuilder.batch from the Java driver and when turning
on auditing. (DSP-3673)
A non-functioning HSHA rpc server (DSP-3675)
Previously NULL values were returned as empty datatypes in hive. Now, the exact information from the
native protocol is returned, except collections, which are returned as empty collections when the value
is NULL. (DSP-3840)
Patches
210
Release notes
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Better error message when adding a collection with the same name than a previously dropped one
(CASSANDRA-6276)
Pig support for hadoop CqlInputFormat (CASSANDRA-6454)
Add inter_dc_stream_throughput_outbound_megabits_per_sec (CASSANDRA-6596)
Fix potential AssertionError with 2ndary indexes (CASSANDRA-6612)
Add option to disable STCS in L0 (CASSANDRA-6621)
Hadoop--Add CqlOutputFormat (CASSANDRA-6927)
Always merge ranges owned by a single node (CASSANDRA-6930)
cqlsh--Wait up to 10 sec for a tracing session (CASSANDRA-7222)
Give CRR a default input_cql Statement (CASSANDRA-7226)
Fix IncompatibleClassChangeError from hadoop2 (CASSANDRA-7229)
Hadoop--allow ACFRW to limit nodes to local DC (CASSANDRA-7252)
Workaround JVM NPE on JMX bind failure (CASSANDRA-7254)
Fix race in FileCacheService RemovalListener (CASSANDRA-7278)
Fix inconsistent use of consistencyForCommit that allowed LOCAL_QUORUM operations to incorrect
become full QUORUM (CASSANDRA-7345)
Make sure high level sstables get compacted (CASSANDRA-7414)
Properly handle unrecognized opcodes and flags (CASSANDRA-7440)
Fix AssertionError when using empty clustering columns and static columns (CASSANDRA-7455)
Hadoop--close CqlRecordWriter clients when finished (CASSANDRA-7459)
Switch liveRatio-related log messages to DEBUG (CASSANDRA-7467)
Set correct stream ID on responses when non-Exception Throwables are thrown while handling native
protocol messages (CASSANDRA-7470)
Remove duplicates from StorageService.getJoiningNodes (CASSANDRA-7478)
Fix error when doing reversed queries with static columns (CASSANDRA-7490)
Properly reject operations on list index with conditions (CASSANDRA-7499)
Throw InvalidRequestException when queries contain relations on entire collection columns
(CASSANDRA-7506)
Don't depend on cassandra config for nodetool ring (CASSANDRA-7508)
Fix truncate to always flush (CASSANDRA-7511)
Warn when SSL certificates have expired (CASSANDRA-7528)
Fix range merging when DES scores are zero (CASSANDRA-7535)
Windows--force range-based repair to non-sequential mode (CASSANDRA-7541)
Fix row size miscalculation in LazilyCompactedRow (CASSANDRA-7543)
Backport CASSNADRA-3569/CASSANDRA-6747 (CASSANDRA-7560)
Remove CqlPagingRecordReader/CqlPagingInputFormat (CASSANDRA-7570)
Fix ReversedType aka DateType mapping to native protocol (CASSANDRA-7576)
cqlsh--enable CTRL-R history search with libedit (CASSANDRA-7577)
Fix sstableloader unable to connect encrypted node (CASSANDRA-7585)
Add stop method to EmbeddedCassandraService (CASSANDRA-7595)
Remove shuffle and taketoken (CASSANDRA-7601)
cqlsh--Add tab-completion for CREATE/DROP USER IF [NOT] EXISTS (CASSANDRA-7611)
Update java driver for hadoop (CASSANDRA-7618)
Fix NPE when listing saved caches dir (CASSANDRA-7632)
Add 'nodetool sethintedhandoffthrottlekb' (CASSANDRA-7635)
cqlsh--cqlsh should automatically disable tracing when selecting from system_traces
(CASSANDRA-7641)
Track max/min timestamps for range tombstones (CASSANDRA-7647)
Add cassandra.auto_bootstrap system property (CASSANDRA-7650)
SimpleSeedProvider no longer caches seeds forever (CASSANDRA-7663)
211
Release notes
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Throw EOFException if we run out of chunks in compressed datafile (CASSANDRA-7664)
Set gc_grace_seconds to seven days for system schema tables (CASSANDRA-7668)
Support connecting to ipv6 jmx with nodetool (CASSANDRA-7669)
Avoid logging CompactionInterrupted at ERROR (CASSANDRA-7694)
Fix potential AssertionError in RangeTombstoneList (CASSANDRA-7700)
cqlsh--Fix failing cqlsh formatting tests (CASSANDRA-7703)
Validate arguments of blobAs functions (CASSANDRA-7707)
Minor leak in sstable2jon (CASSANDRA-7709)
Fix validation when adding static columns (CASSANDRA-7730)
Thrift--fix range deletion of supercolumns (CASSANDRA-7733)
Fix dropping collection when it's the last regular column (CASSANDRA-7744)
Fix race in background compaction check (CASSANDRA-7745)
Do not flush on truncate if durable_writes is false (CASSANDRA-7750)
Fix MS expiring map timeout for Paxos messages (CASSANDRA-7752)
Configure system.paxos with LeveledCompactionStrategy (CASSANDRA-7753)
Fix NPE in FileCacheService.sizeInBytes (CASSANDRA-7756)
Clone token map outside of hot gossip loops (CASSANDRA-7758)
Hadoop--fix cluster initialisation for a split fetching (CASSANDRA-7774)
Fix PRSI handling of CQL3 row markers for row cleanup (CASSANDRA-7787)
Improve PasswordAuthenticator default super user setup (CASSANDRA-7788)
Make StreamReceiveTask thread safe and gc friendly (CASSANDRA-7795)
Stop inheriting liveRatio and liveRatioComputedAt from previous memtables (CASSANDRA-7796)
Validate empty cell names from counter updates (CASSANDRA-7798)
Fix ALTER clustering column type from DateType to TimestampType when using DESC clustering
order (CASSANRDA-7797)
Issues
•
After upgrading DataStax Enterprise from 4.0.0 or 4.0.1 to 4.0.2 or 4.0.3 on RHEL5/CentOS5, the
Snappy JAR file will be missing. (DSP-2558) To restore the Snappy JAR, perform one of the following
procedures to either run the switch-snappy script or re-install DataStax:
•
Run the switch-snappy script:
1. Navigate to the directory containing the switch-snappy script.
$ cd /usr/share/dse ## Package installations
$ cd install_location/bin ## tarball installations
2. Execute the script.
•
$ switch-snappy 1.0.4
To re-install DataStax Enterprise:
1. Uninstall the old installation.
2. Re-install DataStax Enterprise 4.0.4.
•
•
By uninstalling the old installation and re-installing the new one instead of performing an in-place
upgrade, DataStax Enterprise uses the configuration files and data files of the new installation.
DSE Search/Solr cannot index a document that indexes only one field, which is also the unique key in
the schema and the primary key in the corresponding Cassandra table. DSE Search/Solr deletes any
existing data with that primary key and does not return any results for such query. (DSP-3362)
In this release, compaction of files stored on the cfs-archive layer should be disabled, but instead are
compacted automatically. (DSP-4081)
DataStax Enterprise 4.0.3 release notes
DataStax 4.0.3 updates components, and includes improvements, patches, and bug fixes.
212
Release notes
Known issues
•
DataStax Enterprise 4.0.3 and earlier releases are affected by "CVE-2015-0225: Apache Cassandra
remote execution of arbitrary code." Upgrade to the latest release or enable JMX security immediately if
your JMX port is possibly exposed to malicious users.
Components
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Apache Cassandra 2.0.7 (Updated)
Apache Hadoop 1.0.4.10 (Updated)
Apache Hive 0.12.0.2
Apache Pig 0.10.1
Apache Solr 4.6.0.1.3
Apache log4j 1.2.16
Apache Sqoop 1.4.2.14.3
Apache Mahout 0.8
Apache Tomcat 6.0.39 (Updated)
Apache Thrift 0.7.0
Apache Commons
JBCrypt 0.3m
SLF4J 1.7.2
Guava 15.0
JournalIO 1.4.2
Netty 4.0.13.Final
Faster XML 3.1.3
HdrHistogram 1.0.9
Snappy 1.0.5
Cassandra Java Driver 2.0.1
Apache Cassandra documentation covers release notes for Cassandra 2.0.7. NEWS.txt contains latebreaking information about upgrading from previous versions of Cassandra. A NEWS.txt or a NEWS.txt
archive is installed in the following locations:
•
•
Tarball: install_location/resources/cassandra
Package: /usr/share/doc/dse-libcassandra*
NEWS.txt is also posted on the Apache Cassandra project web site.
Enhancements and changes
•
•
•
•
•
Internal authentication support for hadoop, hive, pig, and sqoop commands.
Support for Solr stored copy fields having different source and destination data types.
Enhanced dse and dse-env.sh scripts extend the HADOOP_CLASSPATH of the user and include the
path to Mahout.
Updated Tomcat version 6.0.39 avoids potential garbage collection issues.
In addition to the classic Solr Update Request Processor, in DataStax Enterprise 4.0.3, a custom
version is also available.
Resolved issues
•
Solr
•
•
•
Fixed the issue that caused an error when source and destination copy fields have different
validators. (DSP-1910)
Fixed the issue that caused the DynamicSnitch to break the distributed search when load
information changed during shard selection. (DSP-3322)
Fixed an issue that caused the merging of Solr index segments to fail when using a custom sorting
MergePolicy. (DSP-3230)
213
Release notes
•
•
Using a composite unique key in a Solr schema with a Thrift-compatible table is no longer allowed.
Attempting to load a schema under these conditions results in a message that the operation is not
supported. (DSP-3232)
• Fixed the issue causing copy fields to be applied twice when a Solr HTTP query is inserted data over
CQL 3 tables. (DSP-3240)
• Fixed the issue causing the same row to be indexed several times when you use triggers for copy
fields. (DSP-3241)
Hadoop
•
• Fixed an issue causing an Analytics node to throw an exception endlessly. (DSP-3130)
Other
•
•
•
•
The dsetool rebuild_indexes command now returns an error code when it fails. (DSP-3178)
Using the CQL native protocol, DataStax Enterprise now throws an appropriate exception when
Kerberos authentication issues occur. (DSP-3293)
Fixed the issue breaking CFMetaData. (DSP-3276/CASSANDRA-7074)
Backported a fix to Apache Cassandra to resolve the problem causing a huge performance
regression in tombstone-heavy workloads. (CASSANDRA-6949)
Issues
•
The Solr commitWithin parameter is not supported. (DSP-3021)
DataStax Enterprise 4.0.2 release notes
DataStax Enterprise 4.0 includes updated components, enhancements, changes, resolved issues, and
issues.
Known issues
•
DataStax Enterprise 4.0.2 and earlier releases are affected by "CVE-2015-0225: Apache Cassandra
remote execution of arbitrary code." Upgrade to the latest release or enable JMX security immediately if
your JMX port is possibly exposed to malicious users.
Components
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
214
Apache Cassandra 2.0.6
Apache Hadoop 1.0.4.9
Apache Hive 0.12.0.2
Apache Pig 0.10.1
Apache Solr 4.6.0.1.3
Apache log4j 1.2.16
Apache Sqoop 1.4.2.14.2
Apache Mahout 0.8
Apache Tomcat 6.0.32
Apache Thrift 0.7.0
Apache Commons
JBCrypt 0.3m
SLF4J 1.7.2
Guava 15.0
JournalIO 1.4.2
Netty 4.0.13.Final
Faster XML 3.1.3
HdrHistogram 1.0.9
Snappy 1.0.5
Cassandra Java Driver 2.0.1
Release notes
Enhancements and changes
•
DSE/Search Solr
•
•
•
•
•
•
•
•
DataStax Enterprise 4.0.2 implements a custom version of the OS Solr join query that joins Solr
documents, including those having different Solr cores.
The Solr update log is now officially unsupported because Cassandra provides the functionality.
Configure the update log if you still want to use the update log.
In this release, you can generate Cassandra tracing sessions when running Solr HTTP requests.
DSE Search/Solr now indexes tables that use a CQL composite partition key.
Support for the TimeUUIDTYPE Cassandra validator type has been added.
Support for the EnumField Solr type has also been added.
When you attempt to insert data into Solr fields that are not present in the schema.xml, the exception
thrown by DataStax Enterprise 4.0.2 is more informative than it was in prior releases.
Other
•
•
•
•
In DataStax Enterprise 4.0.2, only JMX ( java management extensions) password authentication is
supported by the dsetool utility. If JMX passwords are enabled, users need to use the passwords
to use the dsetool utility. In earlier releases, Cassandra internal authentication and JMX provided
dsetool password authentication.
A new dsetool command, dsetool status, which is the same as dsetool ring as been added.
The commands list the nodes in the ring including their node type.
If vnodes are enabled on an Analytics or Solr node, an error is logged in the system log and an error
appears in the output of the dsetool ring command.
If the cluster uses a Random or Murmur3 partitioner, the dsetool ring command warns you if
the nodes are imbalanced. The dsetool compares nodes having the most and least load. If the ratio
is greater than 1.1, a warning message appears.
Resolved issues
•
•
•
•
•
•
•
•
•
Fixed an issue causing Solr to always use an external IP address, even though the
EC2MultiRegionSnitch and GossipingPropertyFileSnitch route the traffic on the Cassandra side to
internal IP addresses for nodes in the same data center and to external IP addresses for nodes in
different data centers. This problem resulted in high EC2 bills from Amazon charges for external traffic
and failure in the Rackspace environment, which prevents routing traffic to the external IP address
within the same data center. (DSP-2710)
Fixed the problem with the getFileBlockLocations not returning the correct set of block descriptions.
(DSP-2922)
Fixed an issue causing Solr queries to fail during a short period of time when adding or removing
nodes. (DSP-2975)
Fixed an issue that caused a null pointer exception to be thrown on server shutdown. (DSP-2974)
Fixed issues causing Solr problems working properly with lightweight transactions. (DSP-3028 and
DSP-3032)
Fixed the problem involving the per-segment query results cache to sort queries, such as queries
that do not return a score. Set the useFilterForSortedQuery to true to use the query result cache and
execute the same query multiple times. (DSP-3084)
The problem, which affected the Cassandra commit log, writing to copy fields when those fields were
not stored and when no stored copy fields were present. (DSP-3099)
A row of bad data is no longer inserted through CQL into fields using a copy directive. (DSP-3107)
Fixed the issue preventing batch updates from being indexed if the updates were followed by a delete
and were made on the same partition key. (DSP-3187)
Issues
•
After upgrading DataStax Enterprise from 4.0.0 or 4.0.1 to 4.02 or 4.0.3 on RHEL5/CentOS5, the
Snappy JAR file will be missing. To get it back, either:
215
Release notes
•
Run the switch-snappy script:
$ cd /usr/share/dse ## Package installations
$ cd install_location/bin ## tarball installations
$ switch-snappy 1.0.4
Uninstall the old installation and then do a fresh installation. Using, a regular uninstall maintains the
configuration files and data files.
Cassandra static columns, introduced in Cassandra 2.0.6, cannot be included in the Solr schema (and
hence indexed) for performance reasons because changing the value of a single static column would
require re-indexing all documents sharing the same partition key. (DSP-3143)
•
•
DataStax Enterprise 4.0.1 release notes
Known issues
•
DataStax Enterprise 4.0.1 and earlier releases are affected by "CVE-2015-0225: Apache Cassandra
remote execution of arbitrary code." Upgrade to the latest release or enable JMX security immediately if
your JMX port is possibly exposed to malicious users.
Patches
•
•
A patch to fix the problem with lightweight transactions, compare and set (CAS) that caused Cassandra
to treat an existing row as non-existent if a TTL marker was applied to a column and then the TTL
expired. (CASSANDRA-6623)
A patch to fix the 2.0 HSHA server problem that introduced corrupted data. (CASSANDRA-6285)
Issues
A bug was found that affects upgrades from DSE 4.0.0 on Solr nodes. Upgrades from versions prior to
4.0.0 directly to 4.0.1 are not affected. Please refer to the Upgrade Guide for a detailed workaround.
DataStax Enterprise 4.0 release notes
DataStax Enterprise 4.0 includes updated components, enhancements, changes, resolved issues, and
issues.
Known issues
•
DataStax Enterprise 4.0 is affected by "CVE-2015-0225: Apache Cassandra remote execution of
arbitrary code." Upgrade to the latest release or enable JMX security immediately if your JMX port is
possibly exposed to malicious users.
Components
•
•
•
•
•
•
•
•
•
•
•
Apache Cassandra 2.0.5
Apache Hadoop 1.0.4.9
Apache Hive 0.12.0.1
Apache Pig 0.10.1
Apache Solr 4.6.0.1
Apache log4j 1.2.16
Apache Sqoop 1.4.2.14.1
Apache Mahout 0.8
Apache Tomcat 6.0.32
Apache Thrift 0.7.0
Apache Commons
Apache documentation covers release notes for Cassandra 2.0.5. Cassandra 2.0.5 supports CQL 3.
Enhancements and changes
216
Release notes
DataStax Enterprise 4.0 includes the following enhancements and changes:
•
•
The latest version of the Java SE Runtime Environment (JRE) 7 is required for installing and running
DataStax 4.0.
Support for in-memory tables to accommodate applications, such as ad bidding, that require rapid
response time
Integration of Cassandra 2.0.5, which includes the following features and changes:
•
• Improvements to CQL
• Experimental triggers, configured in CQL, not supported in production deployments
• Performance enhancements
• Column aliases
• Support for accessing legacy CQL tables through cqlsh using the -2 option has been removed
Virtual nodes off by default (differs from Cassandra 2.x)
•
DataStax Enterprise 4.0 turns off virtual nodes (vnodes) by default. DataStax does not recommend
turning on vnodes for Hadoop or Solr nodes, but you can use vnodes for any Cassandra-only cluster, or
a Cassandra-only data center in a mixed Hadoop/Solr/Cassandra deployment. To enable vnodes, see
Using virtual nodes.
Required configuration of initial_token option (differs from Cassandra 2.x)
•
In DataStax Enterprise 4.0, the initial_token option in the default cassandra.yaml of the DataStax
Enterprise-integrated component needs to be set. Because DataStax Enterprise does not use virtual
nodes (vnodes) by default, you need to set the initial_token option.
DSE Search/Solr enhancements and changes:
•
•
•
•
•
•
•
•
•
Optional TCP-based Solr communications that provide lower latency, improved throughput, and
reduced resource consumption
Support for Solr custom field types
New mbeans for obtaining commit and query latency information
Higher maximum calculated heap size limits, 14GB and 10GB, for DSE Search/Solr nodes and
analytics/Hadoop nodes, respectively
Non-string/non-numeric types, such as dates and booleans, can now be used as unique key in the
Solr schema.xml.
Support for Solr 4.3 SolrJ, assuming security is not needed
Solr indexes on CQL 3 tables, including those created using compact storage directive, must now
include the following type mapping version in solrconfig.xml:
<dseTypeMappingVersion>2</dseTypeMappingVersion>.
Hadoop and Hive enhancements:
•
•
•
•
•
•
•
Performance improvements reading MapReduce files
Support for new Hive 0.12 data types: date, varchar, and decimal
Support for the Hive INSERT INTO SELECT statement for copying data from one table and inserting
it into another
• Improved performance of Cassandra File System (CassandraFS) compaction
Changes to command-line options for securing sstableloader data
Lazy loading of compression rules, and in the event of a failure to load, a retry occurs
Configurable disk health checking
Disabled SSTable notifications to improve performance of memtable flushing.
Issues resolved
DataStax Enterprise 4.0 fixes the following issues:
•
•
Fixed an issue causing data to remain after the keyspace and table dropped and recreated.
(CASSANDRA-6635)
Fixed the internal authentication issue when using cassandra-stress. (DSP-2911)
217
Release notes
•
Fixed the problem causing an exception if the PIG_OUTPUTPARTITIONER, which is a
CassandraStorage user-defined function (UDF), is not configured. (DSP-2075)
Issues
•
•
•
GLIBCXX_3.4.9 not found. This error may appear in older Linux distributions when installing DSE from
the binary tarball. The workaround is to replace snappy-java-1.0.5.jar with snappy-java-1.0.4.1.jar.
(DSP-2189)
DataStax Enterprise does not recognize changes to the default log directory used by its Hadoop
component unless you add the HADOOP_LOG_DIR environment variable to the dse-env.sh file, as
described in the Hadoop section of this document.
When the amount of data written to a table exceeds the limit specified by the size_limit_in_mb property,
the following error message is logged:
SEVERE: java.util.concurrent.ExecutionException:
com.datastax.driver.core.exceptions.UnavailableException: Not enough
replica available for query at consistency ONE (1 required but only 0
alive)
•
•
To avoid this condition, manage available memory carefully. (DSP-2990)
In this release, the flush_largest_memtables_at setting is 0.75, which is typically too small causing
excessive flushing of the memtable to disk. The workaround is to change the setting to 0.80 in the
cassandra.yaml. (DSP-2989)
The 2.0 HSHA server introduced corrupted data. This occurs when running the HSHA server in
Cassandra 2.0.x releases earlier than 2.0.6. The hsha (half-synchronous, half-asynchronous) Thrift
server was rewritten on top of Disruptor for Cassandra 2.0. This server can handle more simultaneous
connections than the default sync server. Unfortunately, the rewrite introduced a bug that can cause
incorrect data to be sent from the coordinator to replicas. (CASSANDRA-6285)
Workaround: Use the native protocol or the default sync server. Cassandra 2.0.6 includes the fix
and is expected to be released 3/10/2014. Alternately, you can get the pre-release build from http://
people.apache.org/~slebresne/.
DataStax Enterprise 4.0.1 will include the fix.
218
Tips for using DataStax documentation
Tips for using DataStax documentation
Navigating the documents
To navigate, use the table of contents or search in the left navigation bar. Additional controls are:
Hide or display the left navigation.
Go back or forward through the topics as listed in
the table of contents.
Toggle highlighting of search terms.
Print page.
See doc tweets and provide feedback.
Grab to adjust the size of the navigation pane.
Appears on headings for bookmarking. Right-click
the ¶ to get the link.
Toggles the legend for CQL statements and
nodetool options.
Other resources
You can find more information and help at:
•
•
•
•
•
•
Documentation home page
Datasheets
Webinars
Whitepapers
Developer blogs
Support
219
DataStax Enterprise commons
DataStax Enterprise commons
Common file locations
220
•
•
Packaged installs: /etc/dse/cassandra/cassandra.yaml
Tarball installs: install_location/resources/cassandra/conf/cassandra.yaml
•
•
Packaged installs: /etc/dse/dse.yaml
Tarball installs: install_location/resources/dse/conf/dse.yaml
•
•
Packaged installs: /etc/dse/cassandra/cassandra-topology.properties
Tarball installs: install_location/resources/cassandra/conf/cassandratopology.properties